
import tensorflow as tf
import os
import pandas as pd
from PIL import Image
import matplotlib.pyplot as plt
from typing import Tuple, Dict
import numpy as np
import math

from sklearn.metrics import classification_report, confusion_matrix
from sklearn.utils.class_weight import compute_class_weight
import tensorflow.keras.applications.efficientnet_v2 as effnV2
from tensorflow.keras.layers import Dense, Concatenate, GlobalMaxPooling2D, GlobalAveragePooling2D, Dropout


# Definimos un diccionario para mapear los nombres de las clases con sus índices
dict_map_class = {
    'Loose Silky-bent': 0,
    'Common Chickweed': 1,
    'Scentless Mayweed': 2,
    'Small-flowered Cranesbill': 3,
    'Fat Hen': 4,
    'Charlock': 5,
    'Sugar beet': 6,
    'Cleavers': 7,
    'Black-grass': 8,
    'Shepherds Purse': 9,
    'Common wheat': 10,
    'Maize': 11,   
}

# Invertimos el diccioanrio del mapeo de clases
dict_map_class_inverted = {v: k for k, v in dict_map_class.items()}


def get_dict_dataset(
    dataset_path: str = 'dataset'
):
    """
    Crea un diccionario que asigna las rutas de las imágenes a tuplas que contienen la clase y la partición.

    Args:
        dataset_path (str): Ruta al directorio del conjunto de datos. Por defecto, es 'dataset'.

    Returns:
        dict: Un diccionario donde las claves son las rutas de las imágenes y los valores son tuplas (clase, partición).
    """

    dict_dataset = {}

    train_classes = os.listdir(os.path.join(dataset_path, 'train'))

    for train_class in train_classes:
        class_path = os.path.join(dataset_path, 'train', train_class)
        train_imgs = os.listdir(class_path)
        
        # Split Train images to a 80% for a Train Split for each class
        for train_img in train_imgs[:int(len(train_imgs)*0.8)]:
            train_img_path = os.path.join(class_path, train_img)
            dict_dataset[train_img_path] = (dict_map_class[train_class], 'Train')

        # Assign the rest 20% to Valid Split for each class
        for valid_img in train_imgs[int(len(train_imgs)*0.8):]:
            valid_img_path = os.path.join(class_path, valid_img)
            dict_dataset[valid_img_path] = (dict_map_class[train_class], 'Valid')

    # Geting Test Images
    test_path = os.path.join(dataset_path, 'test')
    test_imgs = os.listdir(test_path)

    for test_img in test_imgs:
        test_img_path = os.path.join(test_path, test_img)
        dict_dataset[test_img_path] = ("Unkown", 'Test')

    return dict_dataset


def dict2dataframe(
    input_dict: Dict[str, Tuple]
):
    """
    Convierte un diccionario a un DataFrame de pandas con columnas para 'path', 'label', y 'split'.

    Args:
        input_dict (dict): Un diccionario donde las claves son las rutas de las imágenes y los valores son tuplas (etiqueta, división).

    Returns:
        pd.DataFrame: Un DataFrame con columnas 'path', 'label' y 'split'.
    """
    df = pd.DataFrame([(key, values[0], values[1]) for key, values in input_dict.items()], columns=['path', 'label', 'split'])

    # Returns shuffled datasets
    return df.sample(frac=1, random_state=42).reset_index(drop=True)


def tf_augmenter():
    """
    Devuelve una función de TensorFlow para la aumentación de datos.

    La función devuelta aplica transformaciones aleatorias a las imágenes de entrada,
    incluyendo volteos aleatorios (arriba-abajo y izquierda-derecha), ajustes de brillo aleatorios,
    y ajustes de contraste aleatorios.

    Returns:
        callable: Una función de TensorFlow que toma un número variable de argumentos (tensores)
                  representando el conjunto de datos y aplica aumentación de datos a las imágenes.
    """
    @tf.function
    def f(*dataset):
        output= list(dataset)
        image = output[0]
        
        # De manera probabilística, con 50% de que ocurra, se aplicará
        # una de estas transformaciones
        if tf.random.uniform([1], minval=0, maxval=1) > 0.5:
            image = tf.image.random_flip_up_down(image)
        if tf.random.uniform([1], minval=0, maxval=1) > 0.5:
            image = tf.image.random_flip_left_right(image)
        if tf.random.uniform([1], minval=0, maxval=1) > 0.5:
            image = tf.image.random_brightness(image, 0.15)
        if tf.random.uniform([1], minval=0, maxval=1) > 0.7:
            image = tf.image.random_contrast(image, 0.6, 1.4)

        output[0] = image
        return output
    return f


@tf.function
def load_image(*inputs):
    """
    Función de TensorFlow para cargar una imagen utilizando una función de NumPy.

    Args:
        *inputs: Número variable de tensores de entrada.

    Returns:
        list: Una lista de tensores de salida con la imagen cargada como primer elemento.
    """
    outputs = list(inputs)
    image = tf.numpy_function(load_image_np, [inputs[0]], tf.float32)
    image.set_shape([None, None, 3])
    outputs[0] = image
    
    return outputs


def load_image_np(path):
    """
    Carga una imagen con PIL desde la ruta especificada y la convierte en un array de NumPy.

    Args:
        path (str): La ruta al archivo de la imagen.

    Returns:
        np.ndarray: Un array de NumPy que representa la imagen cargada en formato RGB.
    """
    return np.array(Image.open(path).convert('RGB')).astype(np.float32)


def resize(index=0, resize_to=None):
    """
    Devuelve una función de TensorFlow para cambiar el tamaño de una imagen en un conjunto de datos.

    Args:
        index (int): Índice del tensor de la imagen en el conjunto de datos. Por defecto, es 0.
        resize_to (tuple, list o None): Tamaño objetivo para cambiar el tamaño. Si es None, no se realiza ningún cambio de tamaño. Por defecto, es None.

    Returns:
        callable: Una función de TensorFlow que cambia el tamaño de la imagen en el conjunto de datos.
    """
    def f(*dataset):
        output = list(dataset)
        # Hacemos resize al input size deseado
        resized_image = tf.image.resize(dataset[index], resize_to)
        # Casteamos de nuevo a uint8 para dejar la imagen entre 0 y 255
        resized_image = tf.cast(resized_image, tf.uint8)
        output[index] = resized_image
        
        return output
    return f


def preprocess_input(index):
    """
    Devuelve una función de TensorFlow para preprocesar una imagen en un conjunto de datos.

    Args:
        index (int): Índice del tensor de la imagen en el conjunto de datos.

    Returns:
        callable: Una función de TensorFlow que preprocesa la imagen en el conjunto de datos.
    """
    @tf.function
    def f(*dataset):
        output = list(dataset)
        image = dataset[index]
        # Convertimos a tf.float32 la imagen
        image = tf.cast(image, tf.float32)
        # Normalizamos la imagen entre 0 y 1
        image = image / 255.
        output[index] = image
        
        return output
    return f


def get_dataset(
    df: pd.DataFrame,
    input_size: Tuple[int, int],
    shuffle: bool = False,
    batch_size: int = None,
    gray_scale: bool = False,
    augmenter: bool = False,
    num_aug: int = None,
    test_set: bool = False
)->tf.data.Dataset:
    """
    Crea un dataset de TensorFlow a partir de un DataFrame.

    Args:
        df (pd.DataFrame): DataFrame que contiene información sobre el conjunto de datos.
        input_size (Tuple[int, int]): Tupla que representa el tamaño objetivo para cambiar el tamaño de las imágenes.
        shuffle (bool): Si se debe barajar el conjunto de datos. Por defecto, es False.
        batch_size (int): Tamaño del lote para el conjunto de datos. Si es None, no se realiza agrupamiento. Por defecto, es None.
        gray_scale (bool): Si se deben convertir las imágenes a escala de grises. Por defecto, es False.
        augmenter (bool): Si se debe aplicar aumentación de datos. Por defecto, es False.
        num_aug (int): Número de aumentaciones a aplicar si augmenter es True. Por defecto, es None.
        test_set (bool): Si el conjunto de datos es un conjunto de prueba. Por defecto, es False.

    Returns:
        tf.data.Dataset: Un conjunto de datos de TensorFlow preparado según las opciones proporcionadas.
    """
    # Imprimir información sobre la distribución de las etiquetas
    print('Number of instances per label: ',
          pd.Series(df['label']).value_counts(), sep='\n')
    print('\nPercentaje of instances per label: ',
          pd.Series(df['label']).value_counts().div(pd.Series(df['label']).shape[0]),
          sep='\n')

    names = np.array(df['path'], dtype=str)

    if not test_set:
        labels = np.array(tf.keras.utils.to_categorical(df['label'], num_classes=12))
    else:
        labels = np.ones(len(names))

    data = names, labels

    # Creamos un dataet de TensorFlow a partir de las rutas y etiquetas
    dataset = tf.data.Dataset.from_tensor_slices(data)

    # Barajamos el dataset
    if shuffle:
        print(' > Shuffle')
        dataset = dataset.shuffle(len(names))

    # Cargamos las imagenes
    dataset = dataset.map(load_image, num_parallel_calls=tf.data.experimental.AUTOTUNE)

    # Aplicamos el resize
    dataset = dataset.map(resize(resize_to=input_size), num_parallel_calls=tf.data.experimental.AUTOTUNE)

    # Aumentamos los datos
    if augmenter:
        print(f' > Augmentamos datos numero {num_aug}')
        if num_aug == 1:
            dataset = dataset.map(tf_augmenter(), num_parallel_calls=tf.data.experimental.AUTOTUNE)

    # Preprocesamos las entradas
    dataset = dataset.map(preprocess_input(0), num_parallel_calls=tf.data.experimental.AUTOTUNE)

    # Convertimos a escala de grises
    if gray_scale:
        print(' > Escala de grises')
        dataset = dataset.map(lambda *args: (tf.image.rgb_to_grayscale(args[0]), *args[1:]))

    # Definimos el batch size
    if batch_size is not None:
        print(' > Establecemos el batchsize')
        dataset = dataset.batch(batch_size)
    
    # Prefetch para paralelizar los siguientes batches mientras se evalua un batch
    dataset = dataset.prefetch(tf.data.experimental.AUTOTUNE)

    return dataset


dict_dataset = get_dict_dataset(dataset_path="dataset/")

# Creamos el DataFrame 
df_dataset = dict2dataframe(dict_dataset)
df_dataset


# Creamos dataset de Train
df_dataset_train = df_dataset[df_dataset['split'] == 'Train'].drop(columns=['split'])
df_dataset_train


# Visualizamos distribuciones por clase
df_dataset_train['label'].value_counts()

label
0     523
1     488
2     412
3     396
4     380
5     312
6     308
7     229
8     210
9     184
10    176
11    176
Name: count, dtype: int64


# Calculamos los pesos de cada clase debido al desbalanceo de las clases
class_weights = compute_class_weight(class_weight='balanced',
                                     classes=np.unique(df_dataset_train['label'].values),
                                     y=df_dataset_train['label'].values
)

class_weight_dict = dict(enumerate(class_weights))

class_weight_dict

{0: 0.6045251752708731,
 1: 0.6478825136612022,
 2: 0.7673948220064725,
 3: 0.7984006734006734,
 4: 0.8320175438596491,
 5: 1.0133547008547008,
 6: 1.0265151515151516,
 7: 1.3806404657933042,
 8: 1.5055555555555555,
 9: 1.7182971014492754,
 10: 1.7964015151515151,
 11: 1.7964015151515151}


# Creamos dataset de Validación
df_dataset_valid = df_dataset[df_dataset['split'] == 'Valid'].drop(columns=['split'])
df_dataset_valid


# Visualizamos distribuciones por clase
df_dataset_valid['label'].value_counts()

label
0     131
1     123
2     104
3     100
4      95
5      78
6      77
7      58
8      53
9      47
11     45
10     45
Name: count, dtype: int64


# Creamos dataset de Test
df_dataset_test = df_dataset[df_dataset['split'] == 'Test'].drop(columns=['split'])
df_dataset_test


# Visualizamos distribuciones por clase
df_dataset_test['label'].value_counts()

label
Unkown    794
Name: count, dtype: int64


fig = plt.figure(figsize = (20, 5))

for i in range(20):
    img = Image.open(df_dataset_train.iloc[i]['path'])
    ax = fig.add_subplot(2, 10, i + 1)
    ax.imshow(img.convert('RGB'))
    title = dict_map_class_inverted[df_dataset_train.iloc[i]['label']]
    ax.set_title(title, fontsize=10)
    ax.axis('off')


# Creamos un ejemplo de datastet de train
train_tfdataset = get_dataset(
    df=df_dataset_train,
    input_size=(224,224),
    batch_size=16,
    shuffle=True,
    gray_scale=False,
    augmenter=True,
    num_aug=1,
)

train_tfdataset

Number of instances per label: 
label
0     523
1     488
2     412
3     396
4     380
5     312
6     308
7     229
8     210
9     184
10    176
11    176
Name: count, dtype: int64

Percentaje of instances per label: 
label
0     0.137849
1     0.128624
2     0.108593
3     0.104375
4     0.100158
5     0.082235
6     0.081181
7     0.060358
8     0.055351
9     0.048498
10    0.046389
11    0.046389
Name: count, dtype: float64

2023-11-26 18:42:44.832637: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:880] could not open file to read NUMA node: /sys/bus/pci/devices/0000:06:00.0/numa_node
Your kernel may have been built without NUMA support.
2023-11-26 18:42:44.837225: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:880] could not open file to read NUMA node: /sys/bus/pci/devices/0000:06:00.0/numa_node
Your kernel may have been built without NUMA support.
2023-11-26 18:42:44.837457: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:880] could not open file to read NUMA node: /sys/bus/pci/devices/0000:06:00.0/numa_node
Your kernel may have been built without NUMA support.
2023-11-26 18:42:44.840091: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:880] could not open file to read NUMA node: /sys/bus/pci/devices/0000:06:00.0/numa_node
Your kernel may have been built without NUMA support.
2023-11-26 18:42:44.840385: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:880] could not open file to read NUMA node: /sys/bus/pci/devices/0000:06:00.0/numa_node
Your kernel may have been built without NUMA support.
2023-11-26 18:42:44.840636: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:880] could not open file to read NUMA node: /sys/bus/pci/devices/0000:06:00.0/numa_node
Your kernel may have been built without NUMA support.
2023-11-26 18:42:45.186309: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:880] could not open file to read NUMA node: /sys/bus/pci/devices/0000:06:00.0/numa_node
Your kernel may have been built without NUMA support.
2023-11-26 18:42:45.186661: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:880] could not open file to read NUMA node: /sys/bus/pci/devices/0000:06:00.0/numa_node
Your kernel may have been built without NUMA support.
2023-11-26 18:42:45.186697: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1977] Could not identify NUMA node of platform GPU id 0, defaulting to 0.  Your kernel may not have been built with NUMA support.
2023-11-26 18:42:45.187050: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:880] could not open file to read NUMA node: /sys/bus/pci/devices/0000:06:00.0/numa_node
Your kernel may have been built without NUMA support.
2023-11-26 18:42:45.187096: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1886] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 7551 MB memory:  -> device: 0, name: NVIDIA GeForce RTX 3080, pci bus id: 0000:06:00.0, compute capability: 8.6

 > Shuffle
 > Augmentamos datos numero 1
 > Establecemos el batchsize

<_PrefetchDataset element_spec=(TensorSpec(shape=(None, 224, 224, 3), dtype=tf.float32, name=None), TensorSpec(shape=(None, 12), dtype=tf.float32, name=None))>


# Visualizamos dataset de Train
batch = next(iter(train_tfdataset))

batches_card_np = batch[0].numpy()
y = batch[1].numpy()
images = list(iter(batches_card_np))

fig, axs = plt.subplots(math.ceil(len(images)/4), 4, figsize=(15, math.ceil(len(images)/4)*4))
axs = axs.ravel()

for i, image in enumerate(images):
    axs[i].imshow(image, cmap='gray')
    title = f"{dict_map_class_inverted[np.argmax(y[i])]}"
    axs[i].set_title(title)


# Creamos un ejemplo de datastet de Validacion
valid_tfdataset = get_dataset(
    df=df_dataset_valid,
    input_size=(224,224),
    batch_size=16,
    shuffle=True,
    gray_scale=False,
    augmenter=False,
)

valid_tfdataset

Number of instances per label: 
label
0     131
1     123
2     104
3     100
4      95
5      78
6      77
7      58
8      53
9      47
11     45
10     45
Name: count, dtype: int64

Percentaje of instances per label: 
label
0     0.137029
1     0.128661
2     0.108787
3     0.104603
4     0.099372
5     0.081590
6     0.080544
7     0.060669
8     0.055439
9     0.049163
11    0.047071
10    0.047071
Name: count, dtype: float64
 > Shuffle
 > Establecemos el batchsize

<_PrefetchDataset element_spec=(TensorSpec(shape=(None, 224, 224, 3), dtype=tf.float32, name=None), TensorSpec(shape=(None, 12), dtype=tf.float32, name=None))>


# Visualizamos dataset de Train
batch = next(iter(valid_tfdataset))

batches_card_np = batch[0].numpy()
y = batch[1].numpy()
images = list(iter(batches_card_np))

fig, axs = plt.subplots(math.ceil(len(images)/4), 4, figsize=(15, math.ceil(len(images)/4)*4))
axs = axs.ravel()

for i, image in enumerate(images):
    axs[i].imshow(image, cmap='gray')
    title = f"{dict_map_class_inverted[np.argmax(y[i])]}"
    axs[i].set_title(title)


# Creamos un ejemplo de datastet de test
test_tfdataset = get_dataset(
    df=df_dataset_test,
    input_size=(224,224),
    batch_size=16,
    gray_scale=False,
    augmenter=False,
    test_set=True,
)

test_tfdataset

Number of instances per label: 
label
Unkown    794
Name: count, dtype: int64

Percentaje of instances per label: 
label
Unkown    1.0
Name: count, dtype: float64
 > Establecemos el batchsize

<_PrefetchDataset element_spec=(TensorSpec(shape=(None, 224, 224, 3), dtype=tf.float32, name=None), TensorSpec(shape=(None,), dtype=tf.float64, name=None))>


# Visualizamos dataset de Test
batch = next(iter(test_tfdataset))

batches_card_np = batch[0].numpy()
y = batch[1].numpy()
images = list(iter(batches_card_np))

fig, axs = plt.subplots(math.ceil(len(images)/4), 4, figsize=(15, math.ceil(len(images)/4)*4))
axs = axs.ravel()

for i, image in enumerate(images):
    axs[i].imshow(image, cmap='gray')
    title = f"{dict_map_class_inverted[np.argmax(y[i])]}"
    axs[i].set_title(title)


from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping
from tensorflow.keras.callbacks import Callback
import matplotlib
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd

class HistoryAndPlotCallback(Callback):
    def __init__(self, model_name):
        super(HistoryAndPlotCallback, self).__init__()
        self.model_name = model_name

    def on_train_end(self, logs=None):
        # Guardar el historial del modelo
        history = self.model.history.history

        # Calcular límite eje Y
        loss_av = np.average(history["loss"][-1])
        val_loss_av = np.average(history["val_loss"][-1])
        acc_av = np.average(history["acc"][-1])
        val_acc_av = np.average(history["val_acc"][-1])
        y_upper_limit = np.max([loss_av, val_loss_av, acc_av, val_acc_av])
        y_upper_limit *= 2

        np.save(f'models/{self.model_name}/training_history.npy', history)

        # Guardar métricas y pérdidas en una sola gráfica
        epochs = len(history["loss"])
        
        plt.style.use("ggplot")
        plt.figure(figsize=(12, 8))
        plt.plot(np.arange(0, epochs), history["loss"], label="train_loss")
        plt.plot(np.arange(0, epochs), history["val_loss"], label="val_loss")
        plt.plot(np.arange(0, epochs), history["acc"], label="train_acc")
        plt.plot(np.arange(0, epochs), history["val_acc"], label="val_acc")
        
        plt.title(f"Training Loss and Accuracy - {self.model_name}")
        plt.xlabel("Epoch #")
        plt.ylabel("Loss/Accuracy")
        plt.legend(loc='upper left')
        plt.ylim(0, int(y_upper_limit))
        plt.grid(True)
        
        # Guardar la gráfica
        plt.savefig(f'models/{self.model_name}/training_plot.png')
        plt.close()

        # Guardar history para revisar en el futuro por si se activa un earlystopping o cualquier problema
        pd.DataFrame.from_dict(history).to_csv(f'models/{self.model_name}/history.csv', index=False)


def get_callbacks(model_name):
    callbacks = [
        EarlyStopping(
            monitor = 'val_loss', 
            mode = 'min', 
            patience = 20, 
            verbose=1,
            restore_best_weights=False),
        ModelCheckpoint(
            filepath = f'models/{model_name}',
            monitor = 'val_loss',
            mode = 'min',
            save_best_only = True,
            verbose = 1),
        HistoryAndPlotCallback(model_name),

    ]
        
    return callbacks


print("Dataset Train (TensorFlow):")

train_tfdataset = get_dataset(
    df=df_dataset_train,
    input_size=(224,224),
    batch_size=32,
    shuffle=True,
    gray_scale=False,
    augmenter=True,
    num_aug=1,
)
train_tfdataset

Dataset Train (TensorFlow):
Number of instances per label: 
label
0     523
1     488
2     412
3     396
4     380
5     312
6     308
7     229
8     210
9     184
10    176
11    176
Name: count, dtype: int64

Percentaje of instances per label: 
label
0     0.137849
1     0.128624
2     0.108593
3     0.104375
4     0.100158
5     0.082235
6     0.081181
7     0.060358
8     0.055351
9     0.048498
10    0.046389
11    0.046389
Name: count, dtype: float64
 > Shuffle
 > Augmentamos datos numero 1
 > Establecemos el batchsize

<_PrefetchDataset element_spec=(TensorSpec(shape=(None, 224, 224, 3), dtype=tf.float32, name=None), TensorSpec(shape=(None, 12), dtype=tf.float32, name=None))>


print("Dataset Valid (TensorFlow):")

valid_tfdataset = get_dataset(
    df=df_dataset_valid,
    input_size=(224,224),
    batch_size=32,
    shuffle=False,
    gray_scale=False,
    augmenter=False,
)

valid_tfdataset

Dataset Valid (TensorFlow):
Number of instances per label: 
label
0     131
1     123
2     104
3     100
4      95
5      78
6      77
7      58
8      53
9      47
11     45
10     45
Name: count, dtype: int64

Percentaje of instances per label: 
label
0     0.137029
1     0.128661
2     0.108787
3     0.104603
4     0.099372
5     0.081590
6     0.080544
7     0.060669
8     0.055439
9     0.049163
11    0.047071
10    0.047071
Name: count, dtype: float64
 > Establecemos el batchsize

<_PrefetchDataset element_spec=(TensorSpec(shape=(None, 224, 224, 3), dtype=tf.float32, name=None), TensorSpec(shape=(None, 12), dtype=tf.float32, name=None))>


print("Dataset Test (TensorFlow):")

test_tfdataset = get_dataset(
    df=df_dataset_test,
    input_size=(224,224),
    batch_size=32,
    gray_scale=False,
    augmenter=False,
    test_set=True,
)

test_tfdataset

Dataset Test (TensorFlow):
Number of instances per label: 
label
Unkown    794
Name: count, dtype: int64

Percentaje of instances per label: 
label
Unkown    1.0
Name: count, dtype: float64
 > Establecemos el batchsize

<_PrefetchDataset element_spec=(TensorSpec(shape=(None, 224, 224, 3), dtype=tf.float32, name=None), TensorSpec(shape=(None,), dtype=tf.float64, name=None))>


# model_factory
from tensorflow.keras import models, layers

class ConvBlockConfig:
    def __init__(self, filters, kernel_size, use_maxpooling=True, use_batchnorm=True, dropout_value=0, name=None):
        """Configuracion de cada bloque con los parametros que vienen del json

        Args:
            filters (_type_): Numero de filtros del bloque (una capa convolucional por bloque)
            kernel_size (_type_): Tamaño del kernel de la capa convolucional
            use_maxpooling (bool, optional): Usar regularizacion MaxPooling. Defaults to True.
            use_batchnorm (bool, optional): Usar regularizacion BatchNorm. Defaults to True.
            dropout_value (int, optional): Usar regularizacion Dropout. Defaults to 0.
            name (_type_, optional): Nombre del bloque. Defaults to None.
        """
        self.filters = filters
        self.kernel_size = kernel_size
        self.use_maxpooling = use_maxpooling
        self.use_batchnorm = use_batchnorm
        self.dropout_value = dropout_value
        self.name = name

def create_conv_block(config, input_shape):
    """Crea los bloques convolucionales con sus regularizaciones (MaxPooling, BatchNorm y Dropout)

    Args:
        config (_type_): Configuracion de cada bloque (viene del json de experimentos)
        input_shape (_type_): Tamaño de entrada del bloque

    Returns:
        _type_: Bloque convolucional
    """
    # Creacion del modelo
    block = models.Sequential(name=config.name)
    
    # CNN input 
    block.add(layers.Conv2D(config.filters, config.kernel_size, activation='relu', padding='same', input_shape=input_shape))
    
    # Regularizacion
    if config.use_batchnorm:
        block.add(layers.BatchNormalization(name=f'{config.name}_batchnorm'))
    
    if config.use_maxpooling:
        block.add(layers.MaxPooling2D((2, 2), name=f'{config.name}_maxpooling'))
    
    if config.dropout_value > 0:
        block.add(layers.Dropout(config.dropout_value))
    
    return block

def create_model_with_configurations(conv_block_configs, input_shape, num_classes):
    """Crea el modelo en base a los bloques que le hemos pasado con la configuracion de los experimentos

    Args:
        conv_block_configs (_type_): Configuracion de los bloques
        input_shape (_type_): Dimensiones de entrada
        num_classes (_type_): Numeros de clases para la capa de salida

    Returns:
        _type_: Modelo
    """
    model = models.Sequential()
    
    # Incluimos la capa de entrada directamente en el modelo principal
    model.add(layers.InputLayer(input_shape=input_shape, name='input'))
    
    # Creacion de los bloques para la extraccion de caracteristicas
    for i, config in enumerate(conv_block_configs, start=1):
        block = create_conv_block(config, input_shape)
        model.add(block)
        input_shape = (input_shape[0]//2, input_shape[1]//2, config.filters)  # Ajustamos input_shape después de MaxPooling
    
    # Creacion del top model para obtener la clase
    model.add(layers.Flatten(name='top_model_flatten'))
    model.add(layers.Dense(128, activation='relu', name='top_model_dense_1'))
    #model.add(layers.Dropout(0.5, name='top_model_dropout')) # Opcional
    model.add(layers.Dense(num_classes, activation='softmax', name='top_model_output'))
    
    return model


def scratch_train(input_shape, experiment_config:dict):
    # Crear modelo, compilarlo y entrenarlo
    for experiment_name, config in experiment_config.items():
        print(f"Running Experiment: {experiment_name}")

        # Configurar el modelo utilizando las configuraciones
        block_configs = []
        for block_conf in config["model"]["conv_block_configs"]:
            block = ConvBlockConfig(block_conf["filters"], 
                                    block_conf["kernel_size"], 
                                    block_conf["use_maxpooling"],
                                    block_conf["use_batchnorm"],
                                    block_conf["dropout_value"],
                                    block_conf["name"])
            block_configs.append(block)
        
        # Crear modelo
        model = create_model_with_configurations(block_configs, input_shape, num_classes)
        
        # Compilar modelo
        model.compile(
            loss = tf.keras.losses.categorical_crossentropy,
            optimizer = tf.keras.optimizers.Adam(0.0001),
            metrics = ['acc']) 
        
        # Crear callbacks
        callbacks = get_callbacks(experiment_name)

        # Mostrar arquitectura
        print(model.summary())

        # Entrenar modelo
        history = model.fit(train_tfdataset.repeat(),
                            validation_data=valid_tfdataset,
                            class_weight = class_weight_dict,
                            epochs = config["training"]["epochs"],
                            steps_per_epoch=config["training"]["steps_per_epoch"],
                            callbacks = callbacks,
                            verbose=1)


# Configuración experimentos
scratch_experiments = {
    "model1": {
        "model": {
            "conv_block_configs": [
                {
                    "filters": 64, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": True, 
                    "dropout_value": 0.1,
                    "name": "block1"
                },
                
                {
                    "filters": 128, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": True, 
                    "dropout_value": 0.1,
                    "name": "block2"
                },
                
                {
                    "filters": 256, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": True, 
                    "dropout_value": 0.1,
                    "name": "block3"
                }
            ]
        },
        "training": {
            "steps_per_epoch": 32,
            "epochs": 50
        }
    },
    
    "model1_nobatchnorm": {
        "model": {
            "conv_block_configs": [
                {
                    "filters": 64, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": False, 
                    "dropout_value": 0.1,
                    "name": "block1"
                },
                {
                    "filters": 128, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": False, 
                    "dropout_value": 0.1,
                    "name": "block2"},
                {
                    "filters": 256, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": False, 
                    "dropout_value": 0.1,
                    "name": "block3"
                }
            ]
        },
        "training": {
            "steps_per_epoch": 32,
            "epochs": 50
        }
    },
    
    "model2": {
        "model": {
            "conv_block_configs": [
                {
                    "filters": 64, 
                    "kernel_size": [5, 5], 
                    "use_maxpooling": True, 
                    "use_batchnorm": True, 
                    "dropout_value": 0.1,
                    "name": "block1"},
                {
                    "filters": 128, 
                    "kernel_size": [5, 5], 
                    "use_maxpooling": True, 
                    "use_batchnorm": True, 
                    "dropout_value": 0.1,
                    "name": "block2"
                }
            ]
        },
        "training": {
            "steps_per_epoch": 32,
            "epochs": 50
        }
    },

    "model2_nobatchnorm": {
        "model": {
            "conv_block_configs": [
                {
                    "filters": 64, 
                    "kernel_size": [5, 5], 
                    "use_maxpooling": True, 
                    "use_batchnorm": False, 
                    "dropout_value": 0.1,
                    "name": "block1"},
                {
                    "filters": 128, 
                    "kernel_size": [5, 5], 
                    "use_maxpooling": True, 
                    "use_batchnorm": False, 
                    "dropout_value": 0.1,
                    "name": "block2"
                }
            ]
        },
        "training": {
            "steps_per_epoch": 32,
            "epochs": 50
        }
    },

    "model3": {
        "model": {
            "conv_block_configs": [
                {
                    "filters": 64, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": True, 
                    "dropout_value": 0.1,
                    "name": "block1"
                },
                {
                    "filters": 128, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": True, 
                    "dropout_value": 0.1,
                    "name": "block2"
                },
                {
                    "filters": 256, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": True, 
                    "dropout_value": 0.1,
                    "name": "block3"
                },
                {
                    "filters": 512, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": True, 
                    "dropout_value": 0.1,
                    "name": "block4"
                }
            ]
        },
        "training": {
            "steps_per_epoch": 32,
            "epochs": 50
        }
    },

    "model3_nobatchnorm": {
        "model": {
            "conv_block_configs": [
                {
                    "filters": 64, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": False, 
                    "dropout_value": 0.1,
                    "name": "block1"
                },
                {
                    "filters": 128, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": False, 
                    "dropout_value": 0.1,
                    "name": "block2"
                },
                {
                    "filters": 256, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": False, 
                    "dropout_value": 0.1,
                    "name": "block3"
                },
                {
                    "filters": 512, 
                    "kernel_size": [3, 3], 
                    "use_maxpooling": True, 
                    "use_batchnorm": False, 
                    "dropout_value": 0.1,
                    "name": "block4"
                }
            ]
        },
        "training": {
            "steps_per_epoch": 32,
            "epochs": 50
        }
    }
  }


# Dimensiones de entrada
input_shape = (224, 224, 3)

# Número de clases
num_classes = df_dataset_train["label"].max() + 1

# Lanzar entrenamiento
scratch_train(input_shape, scratch_experiments)

Running Experiment: model1
Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      2048      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       74368     
                                                                 
 block3 (Sequential)         (None, 28, 28, 256)       296192    
                                                                 
 top_model_flatten (Flatten  (None, 200704)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               25690240  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 26064396 (99.43 MB)
Trainable params: 26063500 (99.42 MB)
Non-trainable params: 896 (3.50 KB)
_________________________________________________________________
None
Epoch 1/50

2023-11-26 18:42:56.542931: E tensorflow/core/grappler/optimizers/meta_optimizer.cc:961] layout failed: INVALID_ARGUMENT: Size of values 0 does not match size of permutation 4 @ fanin shape insequential/block1/dropout/dropout/SelectV2-2-TransposeNHWCToNCHW-LayoutOptimizer
2023-11-26 18:42:57.095623: I tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:442] Loaded cuDNN version 8700
2023-11-26 18:42:57.397182: I tensorflow/tsl/platform/default/subprocess.cc:304] Start cannot spawn child process: No such file or directory
2023-11-26 18:42:57.927327: I tensorflow/tsl/platform/default/subprocess.cc:304] Start cannot spawn child process: No such file or directory
2023-11-26 18:42:58.734794: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7f8ca803a590 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2023-11-26 18:42:58.734834: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): NVIDIA GeForce RTX 3080, Compute Capability 8.6
2023-11-26 18:42:58.739453: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:269] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.
2023-11-26 18:42:58.866988: I ./tensorflow/compiler/jit/device_compiler.h:186] Compiled cluster using XLA!  This line is logged at most once for the lifetime of the process.

 6/32 [====>.........................] - ETA: 2s - loss: 8.9099 - acc: 0.1146WARNING:tensorflow:Callback method `on_train_batch_end` is slow compared to the batch time (batch time: 0.0357s vs `on_train_batch_end` time: 0.0594s). Check your callbacks.
32/32 [==============================] - ETA: 0s - loss: 4.8960 - acc: 0.1680
Epoch 1: val_loss improved from inf to 23.67860, saving model to models/model1
INFO:tensorflow:Assets written to: models/model1/assets

INFO:tensorflow:Assets written to: models/model1/assets

32/32 [==============================] - 18s 380ms/step - loss: 4.8960 - acc: 0.1680 - val_loss: 23.6786 - val_acc: 0.1287
Epoch 2/50
32/32 [==============================] - ETA: 0s - loss: 2.2053 - acc: 0.2080
Epoch 2: val_loss did not improve from 23.67860
32/32 [==============================] - 9s 291ms/step - loss: 2.2053 - acc: 0.2080 - val_loss: 40.6756 - val_acc: 0.1287
Epoch 3/50
32/32 [==============================] - ETA: 0s - loss: 2.1161 - acc: 0.2500
Epoch 3: val_loss did not improve from 23.67860
32/32 [==============================] - 10s 321ms/step - loss: 2.1161 - acc: 0.2500 - val_loss: 50.5102 - val_acc: 0.1287
Epoch 4/50
32/32 [==============================] - ETA: 0s - loss: 1.9974 - acc: 0.2960
Epoch 4: val_loss did not improve from 23.67860
32/32 [==============================] - 10s 315ms/step - loss: 1.9974 - acc: 0.2960 - val_loss: 54.4386 - val_acc: 0.1287
Epoch 5/50
32/32 [==============================] - ETA: 0s - loss: 1.9432 - acc: 0.3125
Epoch 5: val_loss did not improve from 23.67860
32/32 [==============================] - 11s 349ms/step - loss: 1.9432 - acc: 0.3125 - val_loss: 57.2173 - val_acc: 0.1287
Epoch 6/50
32/32 [==============================] - ETA: 0s - loss: 1.7876 - acc: 0.3555
Epoch 6: val_loss did not improve from 23.67860
32/32 [==============================] - 9s 289ms/step - loss: 1.7876 - acc: 0.3555 - val_loss: 51.3227 - val_acc: 0.1287
Epoch 7/50
32/32 [==============================] - ETA: 0s - loss: 1.8529 - acc: 0.3701
Epoch 7: val_loss did not improve from 23.67860
32/32 [==============================] - 10s 323ms/step - loss: 1.8529 - acc: 0.3701 - val_loss: 51.2696 - val_acc: 0.1287
Epoch 8/50
32/32 [==============================] - ETA: 0s - loss: 1.8377 - acc: 0.3554
Epoch 8: val_loss did not improve from 23.67860
32/32 [==============================] - 10s 304ms/step - loss: 1.8377 - acc: 0.3554 - val_loss: 41.3990 - val_acc: 0.1287
Epoch 9/50
32/32 [==============================] - ETA: 0s - loss: 1.6229 - acc: 0.4189
Epoch 9: val_loss did not improve from 23.67860
32/32 [==============================] - 9s 281ms/step - loss: 1.6229 - acc: 0.4189 - val_loss: 34.6635 - val_acc: 0.1287
Epoch 10/50
32/32 [==============================] - ETA: 0s - loss: 1.6856 - acc: 0.4561
Epoch 10: val_loss improved from 23.67860 to 20.57506, saving model to models/model1
INFO:tensorflow:Assets written to: models/model1/assets

INFO:tensorflow:Assets written to: models/model1/assets

32/32 [==============================] - 12s 388ms/step - loss: 1.6856 - acc: 0.4561 - val_loss: 20.5751 - val_acc: 0.1308
Epoch 11/50
32/32 [==============================] - ETA: 0s - loss: 1.7020 - acc: 0.4141
Epoch 11: val_loss improved from 20.57506 to 11.84752, saving model to models/model1
INFO:tensorflow:Assets written to: models/model1/assets

INFO:tensorflow:Assets written to: models/model1/assets

32/32 [==============================] - 12s 371ms/step - loss: 1.7020 - acc: 0.4141 - val_loss: 11.8475 - val_acc: 0.1444
Epoch 12/50
32/32 [==============================] - ETA: 0s - loss: 1.4994 - acc: 0.4762
Epoch 12: val_loss improved from 11.84752 to 3.79100, saving model to models/model1
INFO:tensorflow:Assets written to: models/model1/assets

INFO:tensorflow:Assets written to: models/model1/assets


# Classification report
def print_classification_report(true_labels, predictions):
    predicted_classes = np.argmax(predictions, axis=1)
    report = classification_report(true_labels.tolist(), predicted_classes.tolist())
    
    print("Classification Report:\n", report)


# Cargar arquitecturas de los modelos para cargar pesos en memoria
models_list = []
for experiment_name, config in scratch_experiments.items():
    print(f"Running Experiment: {experiment_name}")

    # Configurar el modelo utilizando las configuraciones
    block_configs = []
    for block_conf in config["model"]["conv_block_configs"]:
        block = ConvBlockConfig(block_conf["filters"], 
                                block_conf["kernel_size"], 
                                block_conf["use_maxpooling"],
                                block_conf["use_batchnorm"],
                                block_conf["dropout_value"],
                                block_conf["name"])
        block_configs.append(block)
    
    # Crear modelo
    model = create_model_with_configurations(block_configs, input_shape, num_classes)
    
    # Compilar modelo
    model.compile(
        loss = tf.keras.losses.categorical_crossentropy,
        optimizer = tf.keras.optimizers.Adam(0.0001),
        metrics = ['acc']) 
    
    # Crear callbacks
    callbacks = get_callbacks(experiment_name)

    # Mostrar arquitectura
    print(model.summary())

    models_list.append(model)

Running Experiment: model1
Model: "sequential_6"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      2048      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       74368     
                                                                 
 block3 (Sequential)         (None, 28, 28, 256)       296192    
                                                                 
 top_model_flatten (Flatten  (None, 200704)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               25690240  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 26064396 (99.43 MB)
Trainable params: 26063500 (99.42 MB)
Non-trainable params: 896 (3.50 KB)
_________________________________________________________________
None
Running Experiment: model1_nobatchnorm
Model: "sequential_7"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      1792      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       73856     
                                                                 
 block3 (Sequential)         (None, 28, 28, 256)       295168    
                                                                 
 top_model_flatten (Flatten  (None, 200704)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               25690240  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 26062604 (99.42 MB)
Trainable params: 26062604 (99.42 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________
None
Running Experiment: model2
Model: "sequential_8"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      5120      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       205440    
                                                                 
 top_model_flatten (Flatten  (None, 401408)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               51380352  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 51592460 (196.81 MB)
Trainable params: 51592076 (196.81 MB)
Non-trainable params: 384 (1.50 KB)
_________________________________________________________________
None
Running Experiment: model2_nobatchnorm
Model: "sequential_9"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      4864      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       204928    
                                                                 
 top_model_flatten (Flatten  (None, 401408)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               51380352  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 51591692 (196.81 MB)
Trainable params: 51591692 (196.81 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________
None
Running Experiment: model3
Model: "sequential_10"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      2048      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       74368     
                                                                 
 block3 (Sequential)         (None, 28, 28, 256)       296192    
                                                                 
 block4 (Sequential)         (None, 14, 14, 512)       1182208   
                                                                 
 top_model_flatten (Flatten  (None, 100352)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               12845184  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 14401548 (54.94 MB)
Trainable params: 14399628 (54.93 MB)
Non-trainable params: 1920 (7.50 KB)
_________________________________________________________________
None
Running Experiment: model3_nobatchnorm
Model: "sequential_11"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      1792      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       73856     
                                                                 
 block3 (Sequential)         (None, 28, 28, 256)       295168    
                                                                 
 block4 (Sequential)         (None, 14, 14, 512)       1180160   
                                                                 
 top_model_flatten (Flatten  (None, 100352)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               12845184  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 14397708 (54.92 MB)
Trainable params: 14397708 (54.92 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________
None


# Cargar modelos en memoria

checkpoint_paths = [
    "models/model1",
    "models/model1_nobatchnorm",
    "models/model2",
    "models/model2_nobatchnorm",
    "models/model3",
    "models/model3_nobatchnorm",
]

predictions = []
for model, checkpoint_filepath in zip(models_list, checkpoint_paths):
    model.load_weights(checkpoint_filepath)
    
    model.summary()
    predictions.append(model.predict(valid_tfdataset))

Model: "sequential_6"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      2048      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       74368     
                                                                 
 block3 (Sequential)         (None, 28, 28, 256)       296192    
                                                                 
 top_model_flatten (Flatten  (None, 200704)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               25690240  
                                                                 
 top_model_output (Dense)    (None, 12)                1548

2023-11-26 19:34:55.980413: W tensorflow/core/util/tensor_slice_reader.cc:98] Could not open models/model1: FAILED_PRECONDITION: models/model1; Is a directory: perhaps your file is in a different file format and you need to use a different restore operator?

                                                                 
=================================================================
Total params: 26064396 (99.43 MB)
Trainable params: 26063500 (99.42 MB)
Non-trainable params: 896 (3.50 KB)
_________________________________________________________________
30/30 [==============================] - 5s 159ms/step
Model: "sequential_7"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      1792      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       73856     
                                                                 
 block3 (Sequential)         (None, 28, 28, 256)       295168    
                                                                 
 top_model_flatten (Flatten  (None, 200704)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               25690240  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 26062604 (99.42 MB)
Trainable params: 26062604 (99.42 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________

2023-11-26 19:35:01.114545: W tensorflow/core/util/tensor_slice_reader.cc:98] Could not open models/model1_nobatchnorm: FAILED_PRECONDITION: models/model1_nobatchnorm; Is a directory: perhaps your file is in a different file format and you need to use a different restore operator?

30/30 [==============================] - 5s 148ms/step

2023-11-26 19:35:05.855373: W tensorflow/core/util/tensor_slice_reader.cc:98] Could not open models/model2: FAILED_PRECONDITION: models/model2; Is a directory: perhaps your file is in a different file format and you need to use a different restore operator?

Model: "sequential_8"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      5120      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       205440    
                                                                 
 top_model_flatten (Flatten  (None, 401408)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               51380352  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 51592460 (196.81 MB)
Trainable params: 51592076 (196.81 MB)
Non-trainable params: 384 (1.50 KB)
_________________________________________________________________
30/30 [==============================] - 5s 148ms/step

2023-11-26 19:35:10.757952: W tensorflow/core/util/tensor_slice_reader.cc:98] Could not open models/model2_nobatchnorm: FAILED_PRECONDITION: models/model2_nobatchnorm; Is a directory: perhaps your file is in a different file format and you need to use a different restore operator?

Model: "sequential_9"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      4864      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       204928    
                                                                 
 top_model_flatten (Flatten  (None, 401408)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               51380352  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 51591692 (196.81 MB)
Trainable params: 51591692 (196.81 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________
30/30 [==============================] - 5s 148ms/step
Model: "sequential_10"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 block1 (Sequential)         (None, 112, 112, 64)      2048      
                                                                 
 block2 (Sequential)         (None, 56, 56, 128)       74368     
                                                                 
 block3 (Sequential)         (None, 28, 28, 256)       296192    
                                                                 
 block4 (Sequential)         (None, 14, 14, 512)       1182208   
                                                                 
 top_model_flatten (Flatten  (None, 100352)            0         
 )                                                               
                                                                 
 top_model_dense_1 (Dense)   (None, 128)               12845184  
                                                                 
 top_model_output (Dense)    (None, 12)                1548      
                                                                 
=================================================================
Total params: 14401548 (54.94 MB)
Trainable params: 14399628 (54.93 MB)
Non-trainable params: 1920 (7.50 KB)
_________________________________________________________________

2023-11-26 19:35:15.647254: W tensorflow/core/util/tensor_slice_reader.cc:98] Could not open models/model3: FAILED_PRECONDITION: models/model3; Is a directory: perhaps your file is in a different file format and you need to use a different restore operator?


# Mostrar classification report
for prediction, checkpoint_filepath in zip(predictions, checkpoint_paths):
    print(f"\nModel: {checkpoint_filepath}")
    print_classification_report(df_dataset_valid["label"].to_numpy(), prediction)

Model: models/model1
Classification Report:
               precision    recall  f1-score   support

           0       0.75      0.55      0.63       131
           1       0.89      0.76      0.82       123
           2       0.88      0.49      0.63       104
           3       0.90      0.84      0.87       100
           4       0.69      0.78      0.73        95
           5       0.81      0.77      0.79        78
           6       0.76      0.79      0.78        77
           7       0.79      0.93      0.86        58
           8       0.33      0.32      0.32        53
           9       0.45      0.83      0.58        47
          10       0.40      0.67      0.50        45
          11       0.72      0.96      0.82        45

    accuracy                           0.71       956
   macro avg       0.70      0.72      0.69       956
weighted avg       0.74      0.71      0.71       956


Model: models/model1_nobatchnorm
Classification Report:
               precision    recall  f1-score   support

           0       0.77      0.69      0.73       131
           1       0.76      0.86      0.81       123
           2       0.91      0.75      0.82       104
           3       0.79      0.87      0.83       100
           4       0.61      0.68      0.64        95
           5       0.89      0.82      0.85        78
           6       0.83      0.70      0.76        77
           7       0.74      0.88      0.80        58
           8       0.52      0.42      0.46        53
           9       0.60      0.55      0.58        47
          10       0.48      0.64      0.55        45
          11       0.76      0.76      0.76        45

    accuracy                           0.74       956
   macro avg       0.72      0.72      0.72       956
weighted avg       0.75      0.74      0.74       956


Model: models/model2
Classification Report:
               precision    recall  f1-score   support

           0       0.57      0.74      0.64       131
           1       0.84      0.52      0.64       123
           2       0.55      0.33      0.41       104
           3       0.80      0.81      0.81       100
           4       0.22      0.39      0.28        95
           5       0.69      0.62      0.65        78
           6       0.81      0.45      0.58        77
           7       0.64      0.74      0.69        58
           8       0.45      0.38      0.41        53
           9       0.40      0.17      0.24        47
          10       0.43      0.58      0.49        45
          11       0.44      0.76      0.56        45

    accuracy                           0.55       956
   macro avg       0.57      0.54      0.53       956
weighted avg       0.60      0.55      0.55       956


Model: models/model2_nobatchnorm
Classification Report:
               precision    recall  f1-score   support

           0       0.72      0.75      0.73       131
           1       0.90      0.77      0.83       123
           2       0.56      0.81      0.66       104
           3       0.85      0.87      0.86       100
           4       0.72      0.53      0.61        95
           5       0.81      0.86      0.83        78
           6       0.73      0.64      0.68        77
           7       0.80      0.84      0.82        58
           8       0.45      0.28      0.35        53
           9       0.61      0.47      0.53        47
          10       0.37      0.56      0.45        45
          11       0.72      0.76      0.74        45

    accuracy                           0.71       956
   macro avg       0.69      0.68      0.67       956
weighted avg       0.72      0.71      0.70       956


Model: models/model3
Classification Report:
               precision    recall  f1-score   support

           0       0.86      0.68      0.76       131
           1       0.95      0.93      0.94       123
           2       0.84      0.86      0.85       104
           3       0.92      0.91      0.91       100
           4       0.89      0.89      0.89        95
           5       0.93      0.86      0.89        78
           6       0.93      0.83      0.88        77
           7       0.86      0.88      0.87        58
           8       0.47      0.53      0.50        53
           9       0.65      0.87      0.75        47
          10       0.63      0.84      0.72        45
          11       0.84      0.93      0.88        45

    accuracy                           0.84       956
   macro avg       0.82      0.83      0.82       956
weighted avg       0.85      0.84      0.84       956


Model: models/model3_nobatchnorm
Classification Report:
               precision    recall  f1-score   support

           0       0.82      0.59      0.68       131
           1       0.86      0.90      0.88       123
           2       0.73      0.90      0.81       104
           3       0.95      0.88      0.91       100
           4       0.68      0.76      0.72        95
           5       0.93      0.81      0.86        78
           6       0.86      0.77      0.81        77
           7       0.76      0.91      0.83        58
           8       0.40      0.40      0.40        53
           9       0.76      0.68      0.72        47
          10       0.57      0.71      0.63        45
          11       0.79      0.84      0.82        45

    accuracy                           0.77       956
   macro avg       0.76      0.76      0.76       956
weighted avg       0.78      0.77      0.77       956


from tensorflow.keras.applications import DenseNet121
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Flatten

# Tamaño entrada
input_shape = (224, 224, 3)

# Callbacks y guardado de grádicas
callbacks = get_callbacks("DenseNet121")

    
# Cargar el modelo preentrenado de DenseNet121 sin incluir las capas densas (fully connected)
base_model = DenseNet121(weights='imagenet', include_top=False, input_shape=input_shape)


# Congelar las capas preentrenadas
for layer in base_model.layers:
    layer.trainable = False

# Crear un nuevo modelo Sequential
model = Sequential()

# Agregar la base preentrenada de DenseNet121
model.add(base_model)

# Añadir capas adicionales para la clasificación
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dense(12, activation='softmax'))  

# Compilar el modelo
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['acc'])

# Mostrar la arquitectura del modelo
model.summary()

history = model.fit(
    train_tfdataset.repeat(),
    steps_per_epoch = 32,
    epochs = 50,  
    validation_data = valid_tfdataset,
    callbacks = callbacks
    )

Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/densenet/densenet121_weights_tf_dim_ordering_tf_kernels_notop.h5
29084464/29084464 [==============================] - 2s 0us/step
Model: "sequential_12"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 densenet121 (Functional)    (None, 7, 7, 1024)        7037504   
                                                                 
 flatten (Flatten)           (None, 50176)             0         
                                                                 
 dense (Dense)               (None, 128)               6422656   
                                                                 
 dense_1 (Dense)             (None, 12)                1548      
                                                                 
=================================================================
Total params: 13461708 (51.35 MB)
Trainable params: 6424204 (24.51 MB)
Non-trainable params: 7037504 (26.85 MB)
_________________________________________________________________
Epoch 1/50
32/32 [==============================] - ETA: 0s - loss: 5.0648 - acc: 0.2617
Epoch 1: val_loss improved from inf to 1.84584, saving model to models/DenseNet121
INFO:tensorflow:Assets written to: models/DenseNet121/assets

INFO:tensorflow:Assets written to: models/DenseNet121/assets

32/32 [==============================] - 43s 1s/step - loss: 5.0648 - acc: 0.2617 - val_loss: 1.8458 - val_acc: 0.3944
Epoch 2/50
32/32 [==============================] - ETA: 0s - loss: 1.6440 - acc: 0.4434
Epoch 2: val_loss improved from 1.84584 to 1.34165, saving model to models/DenseNet121
INFO:tensorflow:Assets written to: models/DenseNet121/assets

INFO:tensorflow:Assets written to: models/DenseNet121/assets

32/32 [==============================] - 39s 1s/step - loss: 1.6440 - acc: 0.4434 - val_loss: 1.3416 - val_acc: 0.5429
Epoch 3/50
32/32 [==============================] - ETA: 0s - loss: 1.2199 - acc: 0.6094
Epoch 3: val_loss improved from 1.34165 to 1.31440, saving model to models/DenseNet121
INFO:tensorflow:Assets written to: models/DenseNet121/assets

INFO:tensorflow:Assets written to: models/DenseNet121/assets

32/32 [==============================] - 37s 1s/step - loss: 1.2199 - acc: 0.6094 - val_loss: 1.3144 - val_acc: 0.5659
Epoch 4/50
32/32 [==============================] - ETA: 0s - loss: 1.0090 - acc: 0.6525
Epoch 4: val_loss improved from 1.31440 to 1.08237, saving model to models/DenseNet121
INFO:tensorflow:Assets written to: models/DenseNet121/assets

INFO:tensorflow:Assets written to: models/DenseNet121/assets

32/32 [==============================] - 35s 1s/step - loss: 1.0090 - acc: 0.6525 - val_loss: 1.0824 - val_acc: 0.6151
Epoch 5/50
32/32 [==============================] - ETA: 0s - loss: 0.8035 - acc: 0.7266
Epoch 5: val_loss improved from 1.08237 to 0.78110, saving model to models/DenseNet121
INFO:tensorflow:Assets written to: models/DenseNet121/assets

INFO:tensorflow:Assets written to: models/DenseNet121/assets


# Tamaño entrada
input_shape = (224, 224, 3)

# Callbacks y guardado de grádicas
callbacks = get_callbacks("DenseNet121")

    
# Cargar el modelo preentrenado de DenseNet121 sin incluir las capas densas (fully connected)
base_model = DenseNet121(weights='imagenet', include_top=False, input_shape=input_shape)


# Congelar las capas preentrenadas
for layer in base_model.layers:
    layer.trainable = False

# Crear un nuevo modelo Sequential
model = Sequential()

# Agregar la base preentrenada de DenseNet121
model.add(base_model)

# Añadir capas adicionales para la clasificación
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dense(12, activation='softmax'))  

# Compilar el modelo
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['acc'])

# Mostrar la arquitectura del modelo
model.summary()

# Sacar prediccion
model.load_weights("models/DenseNet121")
prediction = model.predict(valid_tfdataset)

Model: "sequential_20"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 densenet121 (Functional)    (None, 7, 7, 1024)        7037504   
                                                                 
 flatten_8 (Flatten)         (None, 50176)             0         
                                                                 
 dense_26 (Dense)            (None, 128)               6422656   
                                                                 
 dense_27 (Dense)            (None, 12)                1548      
                                                                 
=================================================================
Total params: 13461708 (51.35 MB)
Trainable params: 6424204 (24.51 MB)
Non-trainable params: 7037504 (26.85 MB)
_________________________________________________________________

2023-11-26 20:36:44.491259: W tensorflow/core/util/tensor_slice_reader.cc:98] Could not open models/DenseNet121: FAILED_PRECONDITION: models/DenseNet121; Is a directory: perhaps your file is in a different file format and you need to use a different restore operator?

30/30 [==============================] - 6s 159ms/step


# Mostrar classification report
print("Classification DenseNet121")
print_classification_report(df_dataset_valid["label"].to_numpy(), prediction)

Classification DenseNet121
Classification Report:
               precision    recall  f1-score   support

           0       0.79      0.87      0.83       131
           1       0.86      0.88      0.87       123
           2       0.82      0.92      0.87       104
           3       0.90      0.94      0.92       100
           4       0.92      0.87      0.90        95
           5       0.91      0.94      0.92        78
           6       0.91      0.88      0.89        77
           7       0.96      0.78      0.86        58
           8       0.55      0.53      0.54        53
           9       0.81      0.62      0.70        47
          10       0.87      0.58      0.69        45
          11       0.77      0.96      0.85        45

    accuracy                           0.84       956
   macro avg       0.84      0.81      0.82       956
weighted avg       0.85      0.84      0.84       956


from tensorflow.keras.applications import VGG16

# Tamaño entrada
input_shape = (224, 224, 3)

# Callbacks y guardado de grádicas
callbacks = get_callbacks("VGG16")

    
# Cargar el modelo preentrenado de DenseNet121 sin incluir las capas densas (fully connected)
base_model = VGG16(weights='imagenet', include_top=False, input_shape=input_shape)


# Congelar las capas preentrenadas
for layer in base_model.layers:
    layer.trainable = False

# Crear un nuevo modelo Sequential
model = Sequential()

# Agregar la base preentrenada de VGG16
model.add(base_model)

# Añadir capas adicionales para la clasificación
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dense(12, activation='softmax'))  

# Compilar el modelo
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['acc'])

# Mostrar la arquitectura del modelo
model.summary()

history = model.fit(
    train_tfdataset.repeat(),
    steps_per_epoch = 32,
    epochs = 50,  
    validation_data = valid_tfdataset,
    callbacks = callbacks
    )

Model: "sequential_13"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 vgg16 (Functional)          (None, 7, 7, 512)         14714688  
                                                                 
 flatten_1 (Flatten)         (None, 25088)             0         
                                                                 
 dense_2 (Dense)             (None, 128)               3211392   
                                                                 
 dense_3 (Dense)             (None, 12)                1548      
                                                                 
=================================================================
Total params: 17927628 (68.39 MB)
Trainable params: 3212940 (12.26 MB)
Non-trainable params: 14714688 (56.13 MB)
_________________________________________________________________
Epoch 1/50
32/32 [==============================] - ETA: 0s - loss: 2.9209 - acc: 0.2148
Epoch 1: val_loss improved from inf to 2.10990, saving model to models/VGG16
INFO:tensorflow:Assets written to: models/VGG16/assets

INFO:tensorflow:Assets written to: models/VGG16/assets

32/32 [==============================] - 14s 380ms/step - loss: 2.9209 - acc: 0.2148 - val_loss: 2.1099 - val_acc: 0.2939
Epoch 2/50
32/32 [==============================] - ETA: 0s - loss: 2.0168 - acc: 0.3096
Epoch 2: val_loss improved from 2.10990 to 1.86994, saving model to models/VGG16
INFO:tensorflow:Assets written to: models/VGG16/assets

INFO:tensorflow:Assets written to: models/VGG16/assets

32/32 [==============================] - 12s 370ms/step - loss: 2.0168 - acc: 0.3096 - val_loss: 1.8699 - val_acc: 0.4153
Epoch 3/50
32/32 [==============================] - ETA: 0s - loss: 1.8000 - acc: 0.4043
Epoch 3: val_loss improved from 1.86994 to 1.81284, saving model to models/VGG16
INFO:tensorflow:Assets written to: models/VGG16/assets

INFO:tensorflow:Assets written to: models/VGG16/assets

32/32 [==============================] - 11s 336ms/step - loss: 1.8000 - acc: 0.4043 - val_loss: 1.8128 - val_acc: 0.3672
Epoch 4/50
32/32 [==============================] - ETA: 0s - loss: 1.6699 - acc: 0.4416
Epoch 4: val_loss improved from 1.81284 to 1.62056, saving model to models/VGG16
INFO:tensorflow:Assets written to: models/VGG16/assets

INFO:tensorflow:Assets written to: models/VGG16/assets

32/32 [==============================] - 10s 311ms/step - loss: 1.6699 - acc: 0.4416 - val_loss: 1.6206 - val_acc: 0.4404
Epoch 5/50
32/32 [==============================] - ETA: 0s - loss: 1.4949 - acc: 0.5039
Epoch 5: val_loss improved from 1.62056 to 1.45303, saving model to models/VGG16
INFO:tensorflow:Assets written to: models/VGG16/assets

INFO:tensorflow:Assets written to: models/VGG16/assets


# Cargar modelos en memoria

# Tamaño entrada
input_shape = (224, 224, 3)

# Callbacks y guardado de grádicas
callbacks = get_callbacks("VGG16")

    
# Cargar el modelo preentrenado de DenseNet121 sin incluir las capas densas (fully connected)
base_model = VGG16(weights='imagenet', include_top=False, input_shape=input_shape)


# Congelar las capas preentrenadas
for layer in base_model.layers:
    layer.trainable = False

# Crear un nuevo modelo Sequential
model = Sequential()

# Agregar la base preentrenada de VGG16
model.add(base_model)

# Añadir capas adicionales para la clasificación
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dense(12, activation='softmax'))  

# Compilar el modelo
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['acc'])

# Sacar prediccion
model.load_weights("models/VGG16")
model.summary()
prediction = model.predict(valid_tfdataset)

Model: "sequential_17"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 vgg16 (Functional)          (None, 7, 7, 512)         14714688  
                                                                 
 flatten_5 (Flatten)         (None, 25088)             0         
                                                                 
 dense_20 (Dense)            (None, 128)               3211392   
                                                                 
 dense_21 (Dense)            (None, 12)                1548      
                                                                 
=================================================================
Total params: 17927628 (68.39 MB)
Trainable params: 3212940 (12.26 MB)
Non-trainable params: 14714688 (56.13 MB)
_________________________________________________________________

2023-11-26 20:33:55.155807: W tensorflow/core/util/tensor_slice_reader.cc:98] Could not open models/VGG16: FAILED_PRECONDITION: models/VGG16; Is a directory: perhaps your file is in a different file format and you need to use a different restore operator?

30/30 [==============================] - 5s 155ms/step


# Mostrar classification report
print("Classification VGG16")
print_classification_report(df_dataset_valid["label"].to_numpy(), prediction)

Classification VGG16
Classification Report:
               precision    recall  f1-score   support

           0       0.66      0.84      0.74       131
           1       0.88      0.78      0.83       123
           2       0.74      0.72      0.73       104
           3       0.92      0.97      0.94       100
           4       0.72      0.72      0.72        95
           5       0.92      0.85      0.88        78
           6       0.80      0.88      0.84        77
           7       0.80      0.74      0.77        58
           8       0.61      0.47      0.53        53
           9       0.54      0.60      0.57        47
          10       0.69      0.56      0.62        45
          11       0.92      0.76      0.83        45

    accuracy                           0.77       956
   macro avg       0.77      0.74      0.75       956
weighted avg       0.77      0.77      0.77       956


def load_efficienNetV2(
    num_classes = int,
    imagenet_weights = bool,
):
    """
    Loads EfficientNetV2 model with customized top layers for a specific number of classes.

    Args:
        num_classes (int): Number of classes for the final classification layer.

    Returns:
        tf.keras.models.Model: An instance of the EfficientNetV2 model with customized top layers.
    """
    if imagenet_weights:
        print(' > Loading iamgenet weights')
        weights_name = 'imagenet'
    else:
        weights_name =None

    redEffi = effnV2.EfficientNetV2B0(
        include_top=False,
        weights=weights_name,
        input_shape=(224, 224, 3),
    )

    x = redEffi.outputs[0]
    x_mean = GlobalAveragePooling2D()(x)
    x_max = GlobalMaxPooling2D()(x)
    x = Concatenate()([x_mean, x_max])
    x = Dropout(0.3)(x)

    x = Dense(num_classes, activation = 'softmax')(x)

    model = tf.keras.models.Model(
            inputs=redEffi.inputs, outputs=[x])

    return model


# Cargamos el modelo de prueba
model = load_efficienNetV2(
    num_classes=12,
    imagenet_weights=True)

model.summary()

model.compile(
    loss = tf.keras.losses.categorical_crossentropy,
    optimizer = tf.keras.optimizers.Adam(0.0001),
    metrics = ['acc']) 

callbacks = get_callbacks("efficintB0_cw_imagenet")

# Entrenamos el modelo
history = model.fit(train_tfdataset.repeat(),
        validation_data=valid_tfdataset,
        class_weight = class_weight_dict,
        epochs = 50,
        steps_per_epoch=50,
        callbacks = callbacks,
        verbose=1)

 > Loading iamgenet weights
Model: "model_2"
__________________________________________________________________________________________________
 Layer (type)                Output Shape                 Param #   Connected to                  
==================================================================================================
 input_3 (InputLayer)        [(None, 224, 224, 3)]        0         []                            
                                                                                                  
 rescaling_2 (Rescaling)     (None, 224, 224, 3)          0         ['input_3[0][0]']             
                                                                                                  
 normalization_2 (Normaliza  (None, 224, 224, 3)          0         ['rescaling_2[0][0]']         
 tion)                                                                                            
                                                                                                  
 stem_conv (Conv2D)          (None, 112, 112, 32)         864       ['normalization_2[0][0]']     
                                                                                                  
 stem_bn (BatchNormalizatio  (None, 112, 112, 32)         128       ['stem_conv[0][0]']           
 n)                                                                                               
                                                                                                  
 stem_activation (Activatio  (None, 112, 112, 32)         0         ['stem_bn[0][0]']             
 n)                                                                                               
                                                                                                  
 block1a_project_conv (Conv  (None, 112, 112, 16)         4608      ['stem_activation[0][0]']     
 2D)                                                                                              
                                                                                                  
 block1a_project_bn (BatchN  (None, 112, 112, 16)         64        ['block1a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block1a_project_activation  (None, 112, 112, 16)         0         ['block1a_project_bn[0][0]']  
  (Activation)                                                                                    
                                                                                                  
 block2a_expand_conv (Conv2  (None, 56, 56, 64)           9216      ['block1a_project_activation[0
 D)                                                                 ][0]']                        
                                                                                                  
 block2a_expand_bn (BatchNo  (None, 56, 56, 64)           256       ['block2a_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block2a_expand_activation   (None, 56, 56, 64)           0         ['block2a_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block2a_project_conv (Conv  (None, 56, 56, 32)           2048      ['block2a_expand_activation[0]
 2D)                                                                [0]']                         
                                                                                                  
 block2a_project_bn (BatchN  (None, 56, 56, 32)           128       ['block2a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block2b_expand_conv (Conv2  (None, 56, 56, 128)          36864     ['block2a_project_bn[0][0]']  
 D)                                                                                               
                                                                                                  
 block2b_expand_bn (BatchNo  (None, 56, 56, 128)          512       ['block2b_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block2b_expand_activation   (None, 56, 56, 128)          0         ['block2b_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block2b_project_conv (Conv  (None, 56, 56, 32)           4096      ['block2b_expand_activation[0]
 2D)                                                                [0]']                         
                                                                                                  
 block2b_project_bn (BatchN  (None, 56, 56, 32)           128       ['block2b_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block2b_drop (Dropout)      (None, 56, 56, 32)           0         ['block2b_project_bn[0][0]']  
                                                                                                  
 block2b_add (Add)           (None, 56, 56, 32)           0         ['block2b_drop[0][0]',        
                                                                     'block2a_project_bn[0][0]']  
                                                                                                  
 block3a_expand_conv (Conv2  (None, 28, 28, 128)          36864     ['block2b_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block3a_expand_bn (BatchNo  (None, 28, 28, 128)          512       ['block3a_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block3a_expand_activation   (None, 28, 28, 128)          0         ['block3a_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block3a_project_conv (Conv  (None, 28, 28, 48)           6144      ['block3a_expand_activation[0]
 2D)                                                                [0]']                         
                                                                                                  
 block3a_project_bn (BatchN  (None, 28, 28, 48)           192       ['block3a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block3b_expand_conv (Conv2  (None, 28, 28, 192)          82944     ['block3a_project_bn[0][0]']  
 D)                                                                                               
                                                                                                  
 block3b_expand_bn (BatchNo  (None, 28, 28, 192)          768       ['block3b_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block3b_expand_activation   (None, 28, 28, 192)          0         ['block3b_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block3b_project_conv (Conv  (None, 28, 28, 48)           9216      ['block3b_expand_activation[0]
 2D)                                                                [0]']                         
                                                                                                  
 block3b_project_bn (BatchN  (None, 28, 28, 48)           192       ['block3b_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block3b_drop (Dropout)      (None, 28, 28, 48)           0         ['block3b_project_bn[0][0]']  
                                                                                                  
 block3b_add (Add)           (None, 28, 28, 48)           0         ['block3b_drop[0][0]',        
                                                                     'block3a_project_bn[0][0]']  
                                                                                                  
 block4a_expand_conv (Conv2  (None, 28, 28, 192)          9216      ['block3b_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block4a_expand_bn (BatchNo  (None, 28, 28, 192)          768       ['block4a_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block4a_expand_activation   (None, 28, 28, 192)          0         ['block4a_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block4a_dwconv2 (Depthwise  (None, 14, 14, 192)          1728      ['block4a_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block4a_bn (BatchNormaliza  (None, 14, 14, 192)          768       ['block4a_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block4a_activation (Activa  (None, 14, 14, 192)          0         ['block4a_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block4a_se_squeeze (Global  (None, 192)                  0         ['block4a_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block4a_se_reshape (Reshap  (None, 1, 1, 192)            0         ['block4a_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block4a_se_reduce (Conv2D)  (None, 1, 1, 12)             2316      ['block4a_se_reshape[0][0]']  
                                                                                                  
 block4a_se_expand (Conv2D)  (None, 1, 1, 192)            2496      ['block4a_se_reduce[0][0]']   
                                                                                                  
 block4a_se_excite (Multipl  (None, 14, 14, 192)          0         ['block4a_activation[0][0]',  
 y)                                                                  'block4a_se_expand[0][0]']   
                                                                                                  
 block4a_project_conv (Conv  (None, 14, 14, 96)           18432     ['block4a_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block4a_project_bn (BatchN  (None, 14, 14, 96)           384       ['block4a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block4b_expand_conv (Conv2  (None, 14, 14, 384)          36864     ['block4a_project_bn[0][0]']  
 D)                                                                                               
                                                                                                  
 block4b_expand_bn (BatchNo  (None, 14, 14, 384)          1536      ['block4b_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block4b_expand_activation   (None, 14, 14, 384)          0         ['block4b_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block4b_dwconv2 (Depthwise  (None, 14, 14, 384)          3456      ['block4b_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block4b_bn (BatchNormaliza  (None, 14, 14, 384)          1536      ['block4b_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block4b_activation (Activa  (None, 14, 14, 384)          0         ['block4b_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block4b_se_squeeze (Global  (None, 384)                  0         ['block4b_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block4b_se_reshape (Reshap  (None, 1, 1, 384)            0         ['block4b_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block4b_se_reduce (Conv2D)  (None, 1, 1, 24)             9240      ['block4b_se_reshape[0][0]']  
                                                                                                  
 block4b_se_expand (Conv2D)  (None, 1, 1, 384)            9600      ['block4b_se_reduce[0][0]']   
                                                                                                  
 block4b_se_excite (Multipl  (None, 14, 14, 384)          0         ['block4b_activation[0][0]',  
 y)                                                                  'block4b_se_expand[0][0]']   
                                                                                                  
 block4b_project_conv (Conv  (None, 14, 14, 96)           36864     ['block4b_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block4b_project_bn (BatchN  (None, 14, 14, 96)           384       ['block4b_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block4b_drop (Dropout)      (None, 14, 14, 96)           0         ['block4b_project_bn[0][0]']  
                                                                                                  
 block4b_add (Add)           (None, 14, 14, 96)           0         ['block4b_drop[0][0]',        
                                                                     'block4a_project_bn[0][0]']  
                                                                                                  
 block4c_expand_conv (Conv2  (None, 14, 14, 384)          36864     ['block4b_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block4c_expand_bn (BatchNo  (None, 14, 14, 384)          1536      ['block4c_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block4c_expand_activation   (None, 14, 14, 384)          0         ['block4c_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block4c_dwconv2 (Depthwise  (None, 14, 14, 384)          3456      ['block4c_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block4c_bn (BatchNormaliza  (None, 14, 14, 384)          1536      ['block4c_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block4c_activation (Activa  (None, 14, 14, 384)          0         ['block4c_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block4c_se_squeeze (Global  (None, 384)                  0         ['block4c_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block4c_se_reshape (Reshap  (None, 1, 1, 384)            0         ['block4c_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block4c_se_reduce (Conv2D)  (None, 1, 1, 24)             9240      ['block4c_se_reshape[0][0]']  
                                                                                                  
 block4c_se_expand (Conv2D)  (None, 1, 1, 384)            9600      ['block4c_se_reduce[0][0]']   
                                                                                                  
 block4c_se_excite (Multipl  (None, 14, 14, 384)          0         ['block4c_activation[0][0]',  
 y)                                                                  'block4c_se_expand[0][0]']   
                                                                                                  
 block4c_project_conv (Conv  (None, 14, 14, 96)           36864     ['block4c_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block4c_project_bn (BatchN  (None, 14, 14, 96)           384       ['block4c_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block4c_drop (Dropout)      (None, 14, 14, 96)           0         ['block4c_project_bn[0][0]']  
                                                                                                  
 block4c_add (Add)           (None, 14, 14, 96)           0         ['block4c_drop[0][0]',        
                                                                     'block4b_add[0][0]']         
                                                                                                  
 block5a_expand_conv (Conv2  (None, 14, 14, 576)          55296     ['block4c_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block5a_expand_bn (BatchNo  (None, 14, 14, 576)          2304      ['block5a_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block5a_expand_activation   (None, 14, 14, 576)          0         ['block5a_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block5a_dwconv2 (Depthwise  (None, 14, 14, 576)          5184      ['block5a_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block5a_bn (BatchNormaliza  (None, 14, 14, 576)          2304      ['block5a_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block5a_activation (Activa  (None, 14, 14, 576)          0         ['block5a_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block5a_se_squeeze (Global  (None, 576)                  0         ['block5a_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block5a_se_reshape (Reshap  (None, 1, 1, 576)            0         ['block5a_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block5a_se_reduce (Conv2D)  (None, 1, 1, 24)             13848     ['block5a_se_reshape[0][0]']  
                                                                                                  
 block5a_se_expand (Conv2D)  (None, 1, 1, 576)            14400     ['block5a_se_reduce[0][0]']   
                                                                                                  
 block5a_se_excite (Multipl  (None, 14, 14, 576)          0         ['block5a_activation[0][0]',  
 y)                                                                  'block5a_se_expand[0][0]']   
                                                                                                  
 block5a_project_conv (Conv  (None, 14, 14, 112)          64512     ['block5a_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block5a_project_bn (BatchN  (None, 14, 14, 112)          448       ['block5a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block5b_expand_conv (Conv2  (None, 14, 14, 672)          75264     ['block5a_project_bn[0][0]']  
 D)                                                                                               
                                                                                                  
 block5b_expand_bn (BatchNo  (None, 14, 14, 672)          2688      ['block5b_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block5b_expand_activation   (None, 14, 14, 672)          0         ['block5b_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block5b_dwconv2 (Depthwise  (None, 14, 14, 672)          6048      ['block5b_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block5b_bn (BatchNormaliza  (None, 14, 14, 672)          2688      ['block5b_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block5b_activation (Activa  (None, 14, 14, 672)          0         ['block5b_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block5b_se_squeeze (Global  (None, 672)                  0         ['block5b_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block5b_se_reshape (Reshap  (None, 1, 1, 672)            0         ['block5b_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block5b_se_reduce (Conv2D)  (None, 1, 1, 28)             18844     ['block5b_se_reshape[0][0]']  
                                                                                                  
 block5b_se_expand (Conv2D)  (None, 1, 1, 672)            19488     ['block5b_se_reduce[0][0]']   
                                                                                                  
 block5b_se_excite (Multipl  (None, 14, 14, 672)          0         ['block5b_activation[0][0]',  
 y)                                                                  'block5b_se_expand[0][0]']   
                                                                                                  
 block5b_project_conv (Conv  (None, 14, 14, 112)          75264     ['block5b_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block5b_project_bn (BatchN  (None, 14, 14, 112)          448       ['block5b_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block5b_drop (Dropout)      (None, 14, 14, 112)          0         ['block5b_project_bn[0][0]']  
                                                                                                  
 block5b_add (Add)           (None, 14, 14, 112)          0         ['block5b_drop[0][0]',        
                                                                     'block5a_project_bn[0][0]']  
                                                                                                  
 block5c_expand_conv (Conv2  (None, 14, 14, 672)          75264     ['block5b_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block5c_expand_bn (BatchNo  (None, 14, 14, 672)          2688      ['block5c_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block5c_expand_activation   (None, 14, 14, 672)          0         ['block5c_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block5c_dwconv2 (Depthwise  (None, 14, 14, 672)          6048      ['block5c_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block5c_bn (BatchNormaliza  (None, 14, 14, 672)          2688      ['block5c_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block5c_activation (Activa  (None, 14, 14, 672)          0         ['block5c_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block5c_se_squeeze (Global  (None, 672)                  0         ['block5c_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block5c_se_reshape (Reshap  (None, 1, 1, 672)            0         ['block5c_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block5c_se_reduce (Conv2D)  (None, 1, 1, 28)             18844     ['block5c_se_reshape[0][0]']  
                                                                                                  
 block5c_se_expand (Conv2D)  (None, 1, 1, 672)            19488     ['block5c_se_reduce[0][0]']   
                                                                                                  
 block5c_se_excite (Multipl  (None, 14, 14, 672)          0         ['block5c_activation[0][0]',  
 y)                                                                  'block5c_se_expand[0][0]']   
                                                                                                  
 block5c_project_conv (Conv  (None, 14, 14, 112)          75264     ['block5c_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block5c_project_bn (BatchN  (None, 14, 14, 112)          448       ['block5c_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block5c_drop (Dropout)      (None, 14, 14, 112)          0         ['block5c_project_bn[0][0]']  
                                                                                                  
 block5c_add (Add)           (None, 14, 14, 112)          0         ['block5c_drop[0][0]',        
                                                                     'block5b_add[0][0]']         
                                                                                                  
 block5d_expand_conv (Conv2  (None, 14, 14, 672)          75264     ['block5c_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block5d_expand_bn (BatchNo  (None, 14, 14, 672)          2688      ['block5d_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block5d_expand_activation   (None, 14, 14, 672)          0         ['block5d_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block5d_dwconv2 (Depthwise  (None, 14, 14, 672)          6048      ['block5d_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block5d_bn (BatchNormaliza  (None, 14, 14, 672)          2688      ['block5d_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block5d_activation (Activa  (None, 14, 14, 672)          0         ['block5d_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block5d_se_squeeze (Global  (None, 672)                  0         ['block5d_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block5d_se_reshape (Reshap  (None, 1, 1, 672)            0         ['block5d_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block5d_se_reduce (Conv2D)  (None, 1, 1, 28)             18844     ['block5d_se_reshape[0][0]']  
                                                                                                  
 block5d_se_expand (Conv2D)  (None, 1, 1, 672)            19488     ['block5d_se_reduce[0][0]']   
                                                                                                  
 block5d_se_excite (Multipl  (None, 14, 14, 672)          0         ['block5d_activation[0][0]',  
 y)                                                                  'block5d_se_expand[0][0]']   
                                                                                                  
 block5d_project_conv (Conv  (None, 14, 14, 112)          75264     ['block5d_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block5d_project_bn (BatchN  (None, 14, 14, 112)          448       ['block5d_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block5d_drop (Dropout)      (None, 14, 14, 112)          0         ['block5d_project_bn[0][0]']  
                                                                                                  
 block5d_add (Add)           (None, 14, 14, 112)          0         ['block5d_drop[0][0]',        
                                                                     'block5c_add[0][0]']         
                                                                                                  
 block5e_expand_conv (Conv2  (None, 14, 14, 672)          75264     ['block5d_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block5e_expand_bn (BatchNo  (None, 14, 14, 672)          2688      ['block5e_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block5e_expand_activation   (None, 14, 14, 672)          0         ['block5e_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block5e_dwconv2 (Depthwise  (None, 14, 14, 672)          6048      ['block5e_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block5e_bn (BatchNormaliza  (None, 14, 14, 672)          2688      ['block5e_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block5e_activation (Activa  (None, 14, 14, 672)          0         ['block5e_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block5e_se_squeeze (Global  (None, 672)                  0         ['block5e_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block5e_se_reshape (Reshap  (None, 1, 1, 672)            0         ['block5e_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block5e_se_reduce (Conv2D)  (None, 1, 1, 28)             18844     ['block5e_se_reshape[0][0]']  
                                                                                                  
 block5e_se_expand (Conv2D)  (None, 1, 1, 672)            19488     ['block5e_se_reduce[0][0]']   
                                                                                                  
 block5e_se_excite (Multipl  (None, 14, 14, 672)          0         ['block5e_activation[0][0]',  
 y)                                                                  'block5e_se_expand[0][0]']   
                                                                                                  
 block5e_project_conv (Conv  (None, 14, 14, 112)          75264     ['block5e_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block5e_project_bn (BatchN  (None, 14, 14, 112)          448       ['block5e_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block5e_drop (Dropout)      (None, 14, 14, 112)          0         ['block5e_project_bn[0][0]']  
                                                                                                  
 block5e_add (Add)           (None, 14, 14, 112)          0         ['block5e_drop[0][0]',        
                                                                     'block5d_add[0][0]']         
                                                                                                  
 block6a_expand_conv (Conv2  (None, 14, 14, 672)          75264     ['block5e_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6a_expand_bn (BatchNo  (None, 14, 14, 672)          2688      ['block6a_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6a_expand_activation   (None, 14, 14, 672)          0         ['block6a_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6a_dwconv2 (Depthwise  (None, 7, 7, 672)            6048      ['block6a_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6a_bn (BatchNormaliza  (None, 7, 7, 672)            2688      ['block6a_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6a_activation (Activa  (None, 7, 7, 672)            0         ['block6a_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6a_se_squeeze (Global  (None, 672)                  0         ['block6a_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6a_se_reshape (Reshap  (None, 1, 1, 672)            0         ['block6a_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6a_se_reduce (Conv2D)  (None, 1, 1, 28)             18844     ['block6a_se_reshape[0][0]']  
                                                                                                  
 block6a_se_expand (Conv2D)  (None, 1, 1, 672)            19488     ['block6a_se_reduce[0][0]']   
                                                                                                  
 block6a_se_excite (Multipl  (None, 7, 7, 672)            0         ['block6a_activation[0][0]',  
 y)                                                                  'block6a_se_expand[0][0]']   
                                                                                                  
 block6a_project_conv (Conv  (None, 7, 7, 192)            129024    ['block6a_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6a_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6b_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6a_project_bn[0][0]']  
 D)                                                                                               
                                                                                                  
 block6b_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6b_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6b_expand_activation   (None, 7, 7, 1152)           0         ['block6b_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6b_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6b_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6b_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6b_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6b_activation (Activa  (None, 7, 7, 1152)           0         ['block6b_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6b_se_squeeze (Global  (None, 1152)                 0         ['block6b_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6b_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6b_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6b_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6b_se_reshape[0][0]']  
                                                                                                  
 block6b_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6b_se_reduce[0][0]']   
                                                                                                  
 block6b_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6b_activation[0][0]',  
 y)                                                                  'block6b_se_expand[0][0]']   
                                                                                                  
 block6b_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6b_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6b_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6b_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6b_drop (Dropout)      (None, 7, 7, 192)            0         ['block6b_project_bn[0][0]']  
                                                                                                  
 block6b_add (Add)           (None, 7, 7, 192)            0         ['block6b_drop[0][0]',        
                                                                     'block6a_project_bn[0][0]']  
                                                                                                  
 block6c_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6b_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6c_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6c_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6c_expand_activation   (None, 7, 7, 1152)           0         ['block6c_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6c_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6c_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6c_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6c_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6c_activation (Activa  (None, 7, 7, 1152)           0         ['block6c_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6c_se_squeeze (Global  (None, 1152)                 0         ['block6c_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6c_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6c_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6c_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6c_se_reshape[0][0]']  
                                                                                                  
 block6c_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6c_se_reduce[0][0]']   
                                                                                                  
 block6c_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6c_activation[0][0]',  
 y)                                                                  'block6c_se_expand[0][0]']   
                                                                                                  
 block6c_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6c_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6c_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6c_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6c_drop (Dropout)      (None, 7, 7, 192)            0         ['block6c_project_bn[0][0]']  
                                                                                                  
 block6c_add (Add)           (None, 7, 7, 192)            0         ['block6c_drop[0][0]',        
                                                                     'block6b_add[0][0]']         
                                                                                                  
 block6d_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6c_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6d_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6d_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6d_expand_activation   (None, 7, 7, 1152)           0         ['block6d_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6d_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6d_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6d_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6d_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6d_activation (Activa  (None, 7, 7, 1152)           0         ['block6d_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6d_se_squeeze (Global  (None, 1152)                 0         ['block6d_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6d_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6d_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6d_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6d_se_reshape[0][0]']  
                                                                                                  
 block6d_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6d_se_reduce[0][0]']   
                                                                                                  
 block6d_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6d_activation[0][0]',  
 y)                                                                  'block6d_se_expand[0][0]']   
                                                                                                  
 block6d_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6d_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6d_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6d_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6d_drop (Dropout)      (None, 7, 7, 192)            0         ['block6d_project_bn[0][0]']  
                                                                                                  
 block6d_add (Add)           (None, 7, 7, 192)            0         ['block6d_drop[0][0]',        
                                                                     'block6c_add[0][0]']         
                                                                                                  
 block6e_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6d_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6e_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6e_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6e_expand_activation   (None, 7, 7, 1152)           0         ['block6e_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6e_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6e_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6e_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6e_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6e_activation (Activa  (None, 7, 7, 1152)           0         ['block6e_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6e_se_squeeze (Global  (None, 1152)                 0         ['block6e_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6e_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6e_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6e_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6e_se_reshape[0][0]']  
                                                                                                  
 block6e_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6e_se_reduce[0][0]']   
                                                                                                  
 block6e_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6e_activation[0][0]',  
 y)                                                                  'block6e_se_expand[0][0]']   
                                                                                                  
 block6e_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6e_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6e_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6e_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6e_drop (Dropout)      (None, 7, 7, 192)            0         ['block6e_project_bn[0][0]']  
                                                                                                  
 block6e_add (Add)           (None, 7, 7, 192)            0         ['block6e_drop[0][0]',        
                                                                     'block6d_add[0][0]']         
                                                                                                  
 block6f_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6e_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6f_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6f_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6f_expand_activation   (None, 7, 7, 1152)           0         ['block6f_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6f_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6f_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6f_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6f_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6f_activation (Activa  (None, 7, 7, 1152)           0         ['block6f_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6f_se_squeeze (Global  (None, 1152)                 0         ['block6f_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6f_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6f_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6f_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6f_se_reshape[0][0]']  
                                                                                                  
 block6f_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6f_se_reduce[0][0]']   
                                                                                                  
 block6f_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6f_activation[0][0]',  
 y)                                                                  'block6f_se_expand[0][0]']   
                                                                                                  
 block6f_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6f_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6f_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6f_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6f_drop (Dropout)      (None, 7, 7, 192)            0         ['block6f_project_bn[0][0]']  
                                                                                                  
 block6f_add (Add)           (None, 7, 7, 192)            0         ['block6f_drop[0][0]',        
                                                                     'block6e_add[0][0]']         
                                                                                                  
 block6g_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6f_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6g_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6g_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6g_expand_activation   (None, 7, 7, 1152)           0         ['block6g_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6g_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6g_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6g_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6g_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6g_activation (Activa  (None, 7, 7, 1152)           0         ['block6g_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6g_se_squeeze (Global  (None, 1152)                 0         ['block6g_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6g_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6g_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6g_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6g_se_reshape[0][0]']  
                                                                                                  
 block6g_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6g_se_reduce[0][0]']   
                                                                                                  
 block6g_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6g_activation[0][0]',  
 y)                                                                  'block6g_se_expand[0][0]']   
                                                                                                  
 block6g_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6g_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6g_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6g_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6g_drop (Dropout)      (None, 7, 7, 192)            0         ['block6g_project_bn[0][0]']  
                                                                                                  
 block6g_add (Add)           (None, 7, 7, 192)            0         ['block6g_drop[0][0]',        
                                                                     'block6f_add[0][0]']         
                                                                                                  
 block6h_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6g_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6h_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6h_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6h_expand_activation   (None, 7, 7, 1152)           0         ['block6h_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6h_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6h_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6h_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6h_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6h_activation (Activa  (None, 7, 7, 1152)           0         ['block6h_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6h_se_squeeze (Global  (None, 1152)                 0         ['block6h_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6h_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6h_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6h_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6h_se_reshape[0][0]']  
                                                                                                  
 block6h_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6h_se_reduce[0][0]']   
                                                                                                  
 block6h_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6h_activation[0][0]',  
 y)                                                                  'block6h_se_expand[0][0]']   
                                                                                                  
 block6h_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6h_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6h_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6h_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6h_drop (Dropout)      (None, 7, 7, 192)            0         ['block6h_project_bn[0][0]']  
                                                                                                  
 block6h_add (Add)           (None, 7, 7, 192)            0         ['block6h_drop[0][0]',        
                                                                     'block6g_add[0][0]']         
                                                                                                  
 top_conv (Conv2D)           (None, 7, 7, 1280)           245760    ['block6h_add[0][0]']         
                                                                                                  
 top_bn (BatchNormalization  (None, 7, 7, 1280)           5120      ['top_conv[0][0]']            
 )                                                                                                
                                                                                                  
 top_activation (Activation  (None, 7, 7, 1280)           0         ['top_bn[0][0]']              
 )                                                                                                
                                                                                                  
 global_average_pooling2d_2  (None, 1280)                 0         ['top_activation[0][0]']      
  (GlobalAveragePooling2D)                                                                        
                                                                                                  
 global_max_pooling2d_2 (Gl  (None, 1280)                 0         ['top_activation[0][0]']      
 obalMaxPooling2D)                                                                                
                                                                                                  
 concatenate_2 (Concatenate  (None, 2560)                 0         ['global_average_pooling2d_2[0
 )                                                                  ][0]',                        
                                                                     'global_max_pooling2d_2[0][0]
                                                                    ']                            
                                                                                                  
 dropout_2 (Dropout)         (None, 2560)                 0         ['concatenate_2[0][0]']       
                                                                                                  
 dense_2 (Dense)             (None, 12)                   30732     ['dropout_2[0][0]']           
                                                                                                  
==================================================================================================
Total params: 5950044 (22.70 MB)
Trainable params: 5889436 (22.47 MB)
Non-trainable params: 60608 (236.75 KB)
__________________________________________________________________________________________________
Epoch 1/50
50/50 [==============================] - ETA: 0s - loss: 4.4812 - acc: 0.1256
Epoch 1: val_loss improved from inf to 3.59153, saving model to models\efficintB0_cw_imagenet
INFO:tensorflow:Assets written to: models\efficintB0_cw_imagenet\assets

INFO:tensorflow:Assets written to: models\efficintB0_cw_imagenet\assets

50/50 [==============================] - 148s 3s/step - loss: 4.4812 - acc: 0.1256 - val_loss: 3.5915 - val_acc: 0.1088
Epoch 2/50
50/50 [==============================] - ETA: 0s - loss: 2.9424 - acc: 0.2544
Epoch 2: val_loss improved from 3.59153 to 2.28200, saving model to models\efficintB0_cw_imagenet
INFO:tensorflow:Assets written to: models\efficintB0_cw_imagenet\assets

INFO:tensorflow:Assets written to: models\efficintB0_cw_imagenet\assets

50/50 [==============================] - 130s 3s/step - loss: 2.9424 - acc: 0.2544 - val_loss: 2.2820 - val_acc: 0.2981
Epoch 3/50
50/50 [==============================] - ETA: 0s - loss: 2.0486 - acc: 0.3897
Epoch 3: val_loss improved from 2.28200 to 1.90023, saving model to models\efficintB0_cw_imagenet
INFO:tensorflow:Assets written to: models\efficintB0_cw_imagenet\assets

INFO:tensorflow:Assets written to: models\efficintB0_cw_imagenet\assets

50/50 [==============================] - 129s 3s/step - loss: 2.0486 - acc: 0.3897 - val_loss: 1.9002 - val_acc: 0.3630
Epoch 4/50
50/50 [==============================] - ETA: 0s - loss: 1.5630 - acc: 0.5250
Epoch 4: val_loss improved from 1.90023 to 0.72259, saving model to models\efficintB0_cw_imagenet
INFO:tensorflow:Assets written to: models\efficintB0_cw_imagenet\assets

INFO:tensorflow:Assets written to: models\efficintB0_cw_imagenet\assets

50/50 [==============================] - 130s 3s/step - loss: 1.5630 - acc: 0.5250 - val_loss: 0.7226 - val_acc: 0.7458
Epoch 5/50
50/50 [==============================] - ETA: 0s - loss: 1.2692 - acc: 0.5984
Epoch 5: val_loss did not improve from 0.72259
50/50 [==============================] - 114s 2s/step - loss: 1.2692 - acc: 0.5984 - val_loss: 1.0895 - val_acc: 0.6318
Epoch 6/50
50/50 [==============================] - ETA: 0s - loss: 0.9674 - acc: 0.6819
Epoch 6: val_loss did not improve from 0.72259
50/50 [==============================] - 115s 2s/step - loss: 0.9674 - acc: 0.6819 - val_loss: 0.7446 - val_acc: 0.7395
Epoch 7/50
50/50 [==============================] - ETA: 0s - loss: 0.8878 - acc: 0.6988
Epoch 7: val_loss did not improve from 0.72259
50/50 [==============================] - 115s 2s/step - loss: 0.8878 - acc: 0.6988 - val_loss: 1.2229 - val_acc: 0.5628
Epoch 8/50
50/50 [==============================] - ETA: 0s - loss: 0.7413 - acc: 0.7478
Epoch 8: val_loss did not improve from 0.72259
50/50 [==============================] - 114s 2s/step - loss: 0.7413 - acc: 0.7478 - val_loss: 0.9181 - val_acc: 0.6977
Epoch 9/50
50/50 [==============================] - ETA: 0s - loss: 0.6625 - acc: 0.7713
Epoch 9: val_loss improved from 0.72259 to 0.64226, saving model to models\efficintB0_cw_imagenet
INFO:tensorflow:Assets written to: models\efficintB0_cw_imagenet\assets

INFO:tensorflow:Assets written to: models\efficintB0_cw_imagenet\assets


# Cargar modelos en memoria
# Cargamos el modelo de prueba
model = load_efficienNetV2(
    num_classes=12,
    imagenet_weights=True)

model.summary()

model.compile(
    loss = tf.keras.losses.categorical_crossentropy,
    optimizer = tf.keras.optimizers.Adam(0.0001),
    metrics = ['acc']) 

# Sacar prediccion
model.load_weights("models/efficintB0_cw_imagenet")
prediction = model.predict(valid_tfdataset)

 > Loading iamgenet weights
Model: "model_9"
__________________________________________________________________________________________________
 Layer (type)                Output Shape                 Param #   Connected to                  
==================================================================================================
 input_18 (InputLayer)       [(None, 224, 224, 3)]        0         []                            
                                                                                                  
 rescaling_12 (Rescaling)    (None, 224, 224, 3)          0         ['input_18[0][0]']            
                                                                                                  
 normalization_12 (Normaliz  (None, 224, 224, 3)          0         ['rescaling_12[0][0]']        
 ation)                                                                                           
                                                                                                  
 stem_conv (Conv2D)          (None, 112, 112, 32)         864       ['normalization_12[0][0]']    
                                                                                                  
 stem_bn (BatchNormalizatio  (None, 112, 112, 32)         128       ['stem_conv[0][0]']           
 n)                                                                                               
                                                                                                  
 stem_activation (Activatio  (None, 112, 112, 32)         0         ['stem_bn[0][0]']             
 n)                                                                                               
                                                                                                  
 block1a_project_conv (Conv  (None, 112, 112, 16)         4608      ['stem_activation[0][0]']     
 2D)                                                                                              
                                                                                                  
 block1a_project_bn (BatchN  (None, 112, 112, 16)         64        ['block1a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block1a_project_activation  (None, 112, 112, 16)         0         ['block1a_project_bn[0][0]']  
  (Activation)                                                                                    
                                                                                                  
 block2a_expand_conv (Conv2  (None, 56, 56, 64)           9216      ['block1a_project_activation[0
 D)                                                                 ][0]']                        
                                                                                                  
 block2a_expand_bn (BatchNo  (None, 56, 56, 64)           256       ['block2a_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block2a_expand_activation   (None, 56, 56, 64)           0         ['block2a_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block2a_project_conv (Conv  (None, 56, 56, 32)           2048      ['block2a_expand_activation[0]
 2D)                                                                [0]']                         
                                                                                                  
 block2a_project_bn (BatchN  (None, 56, 56, 32)           128       ['block2a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block2b_expand_conv (Conv2  (None, 56, 56, 128)          36864     ['block2a_project_bn[0][0]']  
 D)                                                                                               
                                                                                                  
 block2b_expand_bn (BatchNo  (None, 56, 56, 128)          512       ['block2b_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block2b_expand_activation   (None, 56, 56, 128)          0         ['block2b_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block2b_project_conv (Conv  (None, 56, 56, 32)           4096      ['block2b_expand_activation[0]
 2D)                                                                [0]']                         
                                                                                                  
 block2b_project_bn (BatchN  (None, 56, 56, 32)           128       ['block2b_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block2b_drop (Dropout)      (None, 56, 56, 32)           0         ['block2b_project_bn[0][0]']  
                                                                                                  
 block2b_add (Add)           (None, 56, 56, 32)           0         ['block2b_drop[0][0]',        
                                                                     'block2a_project_bn[0][0]']  
                                                                                                  
 block3a_expand_conv (Conv2  (None, 28, 28, 128)          36864     ['block2b_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block3a_expand_bn (BatchNo  (None, 28, 28, 128)          512       ['block3a_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block3a_expand_activation   (None, 28, 28, 128)          0         ['block3a_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block3a_project_conv (Conv  (None, 28, 28, 48)           6144      ['block3a_expand_activation[0]
 2D)                                                                [0]']                         
                                                                                                  
 block3a_project_bn (BatchN  (None, 28, 28, 48)           192       ['block3a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block3b_expand_conv (Conv2  (None, 28, 28, 192)          82944     ['block3a_project_bn[0][0]']  
 D)                                                                                               
                                                                                                  
 block3b_expand_bn (BatchNo  (None, 28, 28, 192)          768       ['block3b_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block3b_expand_activation   (None, 28, 28, 192)          0         ['block3b_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block3b_project_conv (Conv  (None, 28, 28, 48)           9216      ['block3b_expand_activation[0]
 2D)                                                                [0]']                         
                                                                                                  
 block3b_project_bn (BatchN  (None, 28, 28, 48)           192       ['block3b_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block3b_drop (Dropout)      (None, 28, 28, 48)           0         ['block3b_project_bn[0][0]']  
                                                                                                  
 block3b_add (Add)           (None, 28, 28, 48)           0         ['block3b_drop[0][0]',        
                                                                     'block3a_project_bn[0][0]']  
                                                                                                  
 block4a_expand_conv (Conv2  (None, 28, 28, 192)          9216      ['block3b_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block4a_expand_bn (BatchNo  (None, 28, 28, 192)          768       ['block4a_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block4a_expand_activation   (None, 28, 28, 192)          0         ['block4a_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block4a_dwconv2 (Depthwise  (None, 14, 14, 192)          1728      ['block4a_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block4a_bn (BatchNormaliza  (None, 14, 14, 192)          768       ['block4a_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block4a_activation (Activa  (None, 14, 14, 192)          0         ['block4a_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block4a_se_squeeze (Global  (None, 192)                  0         ['block4a_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block4a_se_reshape (Reshap  (None, 1, 1, 192)            0         ['block4a_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block4a_se_reduce (Conv2D)  (None, 1, 1, 12)             2316      ['block4a_se_reshape[0][0]']  
                                                                                                  
 block4a_se_expand (Conv2D)  (None, 1, 1, 192)            2496      ['block4a_se_reduce[0][0]']   
                                                                                                  
 block4a_se_excite (Multipl  (None, 14, 14, 192)          0         ['block4a_activation[0][0]',  
 y)                                                                  'block4a_se_expand[0][0]']   
                                                                                                  
 block4a_project_conv (Conv  (None, 14, 14, 96)           18432     ['block4a_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block4a_project_bn (BatchN  (None, 14, 14, 96)           384       ['block4a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block4b_expand_conv (Conv2  (None, 14, 14, 384)          36864     ['block4a_project_bn[0][0]']  
 D)                                                                                               
                                                                                                  
 block4b_expand_bn (BatchNo  (None, 14, 14, 384)          1536      ['block4b_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block4b_expand_activation   (None, 14, 14, 384)          0         ['block4b_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block4b_dwconv2 (Depthwise  (None, 14, 14, 384)          3456      ['block4b_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block4b_bn (BatchNormaliza  (None, 14, 14, 384)          1536      ['block4b_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block4b_activation (Activa  (None, 14, 14, 384)          0         ['block4b_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block4b_se_squeeze (Global  (None, 384)                  0         ['block4b_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block4b_se_reshape (Reshap  (None, 1, 1, 384)            0         ['block4b_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block4b_se_reduce (Conv2D)  (None, 1, 1, 24)             9240      ['block4b_se_reshape[0][0]']  
                                                                                                  
 block4b_se_expand (Conv2D)  (None, 1, 1, 384)            9600      ['block4b_se_reduce[0][0]']   
                                                                                                  
 block4b_se_excite (Multipl  (None, 14, 14, 384)          0         ['block4b_activation[0][0]',  
 y)                                                                  'block4b_se_expand[0][0]']   
                                                                                                  
 block4b_project_conv (Conv  (None, 14, 14, 96)           36864     ['block4b_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block4b_project_bn (BatchN  (None, 14, 14, 96)           384       ['block4b_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block4b_drop (Dropout)      (None, 14, 14, 96)           0         ['block4b_project_bn[0][0]']  
                                                                                                  
 block4b_add (Add)           (None, 14, 14, 96)           0         ['block4b_drop[0][0]',        
                                                                     'block4a_project_bn[0][0]']  
                                                                                                  
 block4c_expand_conv (Conv2  (None, 14, 14, 384)          36864     ['block4b_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block4c_expand_bn (BatchNo  (None, 14, 14, 384)          1536      ['block4c_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block4c_expand_activation   (None, 14, 14, 384)          0         ['block4c_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block4c_dwconv2 (Depthwise  (None, 14, 14, 384)          3456      ['block4c_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block4c_bn (BatchNormaliza  (None, 14, 14, 384)          1536      ['block4c_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block4c_activation (Activa  (None, 14, 14, 384)          0         ['block4c_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block4c_se_squeeze (Global  (None, 384)                  0         ['block4c_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block4c_se_reshape (Reshap  (None, 1, 1, 384)            0         ['block4c_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block4c_se_reduce (Conv2D)  (None, 1, 1, 24)             9240      ['block4c_se_reshape[0][0]']  
                                                                                                  
 block4c_se_expand (Conv2D)  (None, 1, 1, 384)            9600      ['block4c_se_reduce[0][0]']   
                                                                                                  
 block4c_se_excite (Multipl  (None, 14, 14, 384)          0         ['block4c_activation[0][0]',  
 y)                                                                  'block4c_se_expand[0][0]']   
                                                                                                  
 block4c_project_conv (Conv  (None, 14, 14, 96)           36864     ['block4c_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block4c_project_bn (BatchN  (None, 14, 14, 96)           384       ['block4c_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block4c_drop (Dropout)      (None, 14, 14, 96)           0         ['block4c_project_bn[0][0]']  
                                                                                                  
 block4c_add (Add)           (None, 14, 14, 96)           0         ['block4c_drop[0][0]',        
                                                                     'block4b_add[0][0]']         
                                                                                                  
 block5a_expand_conv (Conv2  (None, 14, 14, 576)          55296     ['block4c_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block5a_expand_bn (BatchNo  (None, 14, 14, 576)          2304      ['block5a_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block5a_expand_activation   (None, 14, 14, 576)          0         ['block5a_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block5a_dwconv2 (Depthwise  (None, 14, 14, 576)          5184      ['block5a_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block5a_bn (BatchNormaliza  (None, 14, 14, 576)          2304      ['block5a_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block5a_activation (Activa  (None, 14, 14, 576)          0         ['block5a_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block5a_se_squeeze (Global  (None, 576)                  0         ['block5a_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block5a_se_reshape (Reshap  (None, 1, 1, 576)            0         ['block5a_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block5a_se_reduce (Conv2D)  (None, 1, 1, 24)             13848     ['block5a_se_reshape[0][0]']  
                                                                                                  
 block5a_se_expand (Conv2D)  (None, 1, 1, 576)            14400     ['block5a_se_reduce[0][0]']   
                                                                                                  
 block5a_se_excite (Multipl  (None, 14, 14, 576)          0         ['block5a_activation[0][0]',  
 y)                                                                  'block5a_se_expand[0][0]']   
                                                                                                  
 block5a_project_conv (Conv  (None, 14, 14, 112)          64512     ['block5a_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block5a_project_bn (BatchN  (None, 14, 14, 112)          448       ['block5a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block5b_expand_conv (Conv2  (None, 14, 14, 672)          75264     ['block5a_project_bn[0][0]']  
 D)                                                                                               
                                                                                                  
 block5b_expand_bn (BatchNo  (None, 14, 14, 672)          2688      ['block5b_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block5b_expand_activation   (None, 14, 14, 672)          0         ['block5b_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block5b_dwconv2 (Depthwise  (None, 14, 14, 672)          6048      ['block5b_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block5b_bn (BatchNormaliza  (None, 14, 14, 672)          2688      ['block5b_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block5b_activation (Activa  (None, 14, 14, 672)          0         ['block5b_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block5b_se_squeeze (Global  (None, 672)                  0         ['block5b_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block5b_se_reshape (Reshap  (None, 1, 1, 672)            0         ['block5b_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block5b_se_reduce (Conv2D)  (None, 1, 1, 28)             18844     ['block5b_se_reshape[0][0]']  
                                                                                                  
 block5b_se_expand (Conv2D)  (None, 1, 1, 672)            19488     ['block5b_se_reduce[0][0]']   
                                                                                                  
 block5b_se_excite (Multipl  (None, 14, 14, 672)          0         ['block5b_activation[0][0]',  
 y)                                                                  'block5b_se_expand[0][0]']   
                                                                                                  
 block5b_project_conv (Conv  (None, 14, 14, 112)          75264     ['block5b_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block5b_project_bn (BatchN  (None, 14, 14, 112)          448       ['block5b_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block5b_drop (Dropout)      (None, 14, 14, 112)          0         ['block5b_project_bn[0][0]']  
                                                                                                  
 block5b_add (Add)           (None, 14, 14, 112)          0         ['block5b_drop[0][0]',        
                                                                     'block5a_project_bn[0][0]']  
                                                                                                  
 block5c_expand_conv (Conv2  (None, 14, 14, 672)          75264     ['block5b_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block5c_expand_bn (BatchNo  (None, 14, 14, 672)          2688      ['block5c_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block5c_expand_activation   (None, 14, 14, 672)          0         ['block5c_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block5c_dwconv2 (Depthwise  (None, 14, 14, 672)          6048      ['block5c_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block5c_bn (BatchNormaliza  (None, 14, 14, 672)          2688      ['block5c_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block5c_activation (Activa  (None, 14, 14, 672)          0         ['block5c_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block5c_se_squeeze (Global  (None, 672)                  0         ['block5c_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block5c_se_reshape (Reshap  (None, 1, 1, 672)            0         ['block5c_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block5c_se_reduce (Conv2D)  (None, 1, 1, 28)             18844     ['block5c_se_reshape[0][0]']  
                                                                                                  
 block5c_se_expand (Conv2D)  (None, 1, 1, 672)            19488     ['block5c_se_reduce[0][0]']   
                                                                                                  
 block5c_se_excite (Multipl  (None, 14, 14, 672)          0         ['block5c_activation[0][0]',  
 y)                                                                  'block5c_se_expand[0][0]']   
                                                                                                  
 block5c_project_conv (Conv  (None, 14, 14, 112)          75264     ['block5c_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block5c_project_bn (BatchN  (None, 14, 14, 112)          448       ['block5c_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block5c_drop (Dropout)      (None, 14, 14, 112)          0         ['block5c_project_bn[0][0]']  
                                                                                                  
 block5c_add (Add)           (None, 14, 14, 112)          0         ['block5c_drop[0][0]',        
                                                                     'block5b_add[0][0]']         
                                                                                                  
 block5d_expand_conv (Conv2  (None, 14, 14, 672)          75264     ['block5c_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block5d_expand_bn (BatchNo  (None, 14, 14, 672)          2688      ['block5d_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block5d_expand_activation   (None, 14, 14, 672)          0         ['block5d_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block5d_dwconv2 (Depthwise  (None, 14, 14, 672)          6048      ['block5d_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block5d_bn (BatchNormaliza  (None, 14, 14, 672)          2688      ['block5d_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block5d_activation (Activa  (None, 14, 14, 672)          0         ['block5d_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block5d_se_squeeze (Global  (None, 672)                  0         ['block5d_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block5d_se_reshape (Reshap  (None, 1, 1, 672)            0         ['block5d_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block5d_se_reduce (Conv2D)  (None, 1, 1, 28)             18844     ['block5d_se_reshape[0][0]']  
                                                                                                  
 block5d_se_expand (Conv2D)  (None, 1, 1, 672)            19488     ['block5d_se_reduce[0][0]']   
                                                                                                  
 block5d_se_excite (Multipl  (None, 14, 14, 672)          0         ['block5d_activation[0][0]',  
 y)                                                                  'block5d_se_expand[0][0]']   
                                                                                                  
 block5d_project_conv (Conv  (None, 14, 14, 112)          75264     ['block5d_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block5d_project_bn (BatchN  (None, 14, 14, 112)          448       ['block5d_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block5d_drop (Dropout)      (None, 14, 14, 112)          0         ['block5d_project_bn[0][0]']  
                                                                                                  
 block5d_add (Add)           (None, 14, 14, 112)          0         ['block5d_drop[0][0]',        
                                                                     'block5c_add[0][0]']         
                                                                                                  
 block5e_expand_conv (Conv2  (None, 14, 14, 672)          75264     ['block5d_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block5e_expand_bn (BatchNo  (None, 14, 14, 672)          2688      ['block5e_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block5e_expand_activation   (None, 14, 14, 672)          0         ['block5e_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block5e_dwconv2 (Depthwise  (None, 14, 14, 672)          6048      ['block5e_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block5e_bn (BatchNormaliza  (None, 14, 14, 672)          2688      ['block5e_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block5e_activation (Activa  (None, 14, 14, 672)          0         ['block5e_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block5e_se_squeeze (Global  (None, 672)                  0         ['block5e_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block5e_se_reshape (Reshap  (None, 1, 1, 672)            0         ['block5e_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block5e_se_reduce (Conv2D)  (None, 1, 1, 28)             18844     ['block5e_se_reshape[0][0]']  
                                                                                                  
 block5e_se_expand (Conv2D)  (None, 1, 1, 672)            19488     ['block5e_se_reduce[0][0]']   
                                                                                                  
 block5e_se_excite (Multipl  (None, 14, 14, 672)          0         ['block5e_activation[0][0]',  
 y)                                                                  'block5e_se_expand[0][0]']   
                                                                                                  
 block5e_project_conv (Conv  (None, 14, 14, 112)          75264     ['block5e_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block5e_project_bn (BatchN  (None, 14, 14, 112)          448       ['block5e_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block5e_drop (Dropout)      (None, 14, 14, 112)          0         ['block5e_project_bn[0][0]']  
                                                                                                  
 block5e_add (Add)           (None, 14, 14, 112)          0         ['block5e_drop[0][0]',        
                                                                     'block5d_add[0][0]']         
                                                                                                  
 block6a_expand_conv (Conv2  (None, 14, 14, 672)          75264     ['block5e_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6a_expand_bn (BatchNo  (None, 14, 14, 672)          2688      ['block6a_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6a_expand_activation   (None, 14, 14, 672)          0         ['block6a_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6a_dwconv2 (Depthwise  (None, 7, 7, 672)            6048      ['block6a_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6a_bn (BatchNormaliza  (None, 7, 7, 672)            2688      ['block6a_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6a_activation (Activa  (None, 7, 7, 672)            0         ['block6a_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6a_se_squeeze (Global  (None, 672)                  0         ['block6a_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6a_se_reshape (Reshap  (None, 1, 1, 672)            0         ['block6a_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6a_se_reduce (Conv2D)  (None, 1, 1, 28)             18844     ['block6a_se_reshape[0][0]']  
                                                                                                  
 block6a_se_expand (Conv2D)  (None, 1, 1, 672)            19488     ['block6a_se_reduce[0][0]']   
                                                                                                  
 block6a_se_excite (Multipl  (None, 7, 7, 672)            0         ['block6a_activation[0][0]',  
 y)                                                                  'block6a_se_expand[0][0]']   
                                                                                                  
 block6a_project_conv (Conv  (None, 7, 7, 192)            129024    ['block6a_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6a_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6a_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6b_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6a_project_bn[0][0]']  
 D)                                                                                               
                                                                                                  
 block6b_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6b_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6b_expand_activation   (None, 7, 7, 1152)           0         ['block6b_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6b_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6b_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6b_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6b_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6b_activation (Activa  (None, 7, 7, 1152)           0         ['block6b_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6b_se_squeeze (Global  (None, 1152)                 0         ['block6b_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6b_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6b_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6b_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6b_se_reshape[0][0]']  
                                                                                                  
 block6b_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6b_se_reduce[0][0]']   
                                                                                                  
 block6b_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6b_activation[0][0]',  
 y)                                                                  'block6b_se_expand[0][0]']   
                                                                                                  
 block6b_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6b_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6b_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6b_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6b_drop (Dropout)      (None, 7, 7, 192)            0         ['block6b_project_bn[0][0]']  
                                                                                                  
 block6b_add (Add)           (None, 7, 7, 192)            0         ['block6b_drop[0][0]',        
                                                                     'block6a_project_bn[0][0]']  
                                                                                                  
 block6c_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6b_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6c_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6c_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6c_expand_activation   (None, 7, 7, 1152)           0         ['block6c_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6c_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6c_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6c_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6c_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6c_activation (Activa  (None, 7, 7, 1152)           0         ['block6c_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6c_se_squeeze (Global  (None, 1152)                 0         ['block6c_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6c_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6c_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6c_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6c_se_reshape[0][0]']  
                                                                                                  
 block6c_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6c_se_reduce[0][0]']   
                                                                                                  
 block6c_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6c_activation[0][0]',  
 y)                                                                  'block6c_se_expand[0][0]']   
                                                                                                  
 block6c_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6c_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6c_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6c_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6c_drop (Dropout)      (None, 7, 7, 192)            0         ['block6c_project_bn[0][0]']  
                                                                                                  
 block6c_add (Add)           (None, 7, 7, 192)            0         ['block6c_drop[0][0]',        
                                                                     'block6b_add[0][0]']         
                                                                                                  
 block6d_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6c_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6d_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6d_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6d_expand_activation   (None, 7, 7, 1152)           0         ['block6d_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6d_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6d_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6d_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6d_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6d_activation (Activa  (None, 7, 7, 1152)           0         ['block6d_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6d_se_squeeze (Global  (None, 1152)                 0         ['block6d_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6d_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6d_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6d_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6d_se_reshape[0][0]']  
                                                                                                  
 block6d_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6d_se_reduce[0][0]']   
                                                                                                  
 block6d_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6d_activation[0][0]',  
 y)                                                                  'block6d_se_expand[0][0]']   
                                                                                                  
 block6d_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6d_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6d_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6d_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6d_drop (Dropout)      (None, 7, 7, 192)            0         ['block6d_project_bn[0][0]']  
                                                                                                  
 block6d_add (Add)           (None, 7, 7, 192)            0         ['block6d_drop[0][0]',        
                                                                     'block6c_add[0][0]']         
                                                                                                  
 block6e_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6d_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6e_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6e_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6e_expand_activation   (None, 7, 7, 1152)           0         ['block6e_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6e_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6e_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6e_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6e_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6e_activation (Activa  (None, 7, 7, 1152)           0         ['block6e_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6e_se_squeeze (Global  (None, 1152)                 0         ['block6e_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6e_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6e_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6e_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6e_se_reshape[0][0]']  
                                                                                                  
 block6e_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6e_se_reduce[0][0]']   
                                                                                                  
 block6e_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6e_activation[0][0]',  
 y)                                                                  'block6e_se_expand[0][0]']   
                                                                                                  
 block6e_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6e_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6e_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6e_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6e_drop (Dropout)      (None, 7, 7, 192)            0         ['block6e_project_bn[0][0]']  
                                                                                                  
 block6e_add (Add)           (None, 7, 7, 192)            0         ['block6e_drop[0][0]',        
                                                                     'block6d_add[0][0]']         
                                                                                                  
 block6f_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6e_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6f_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6f_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6f_expand_activation   (None, 7, 7, 1152)           0         ['block6f_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6f_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6f_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6f_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6f_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6f_activation (Activa  (None, 7, 7, 1152)           0         ['block6f_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6f_se_squeeze (Global  (None, 1152)                 0         ['block6f_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6f_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6f_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6f_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6f_se_reshape[0][0]']  
                                                                                                  
 block6f_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6f_se_reduce[0][0]']   
                                                                                                  
 block6f_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6f_activation[0][0]',  
 y)                                                                  'block6f_se_expand[0][0]']   
                                                                                                  
 block6f_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6f_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6f_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6f_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6f_drop (Dropout)      (None, 7, 7, 192)            0         ['block6f_project_bn[0][0]']  
                                                                                                  
 block6f_add (Add)           (None, 7, 7, 192)            0         ['block6f_drop[0][0]',        
                                                                     'block6e_add[0][0]']         
                                                                                                  
 block6g_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6f_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6g_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6g_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6g_expand_activation   (None, 7, 7, 1152)           0         ['block6g_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6g_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6g_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6g_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6g_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6g_activation (Activa  (None, 7, 7, 1152)           0         ['block6g_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6g_se_squeeze (Global  (None, 1152)                 0         ['block6g_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6g_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6g_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6g_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6g_se_reshape[0][0]']  
                                                                                                  
 block6g_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6g_se_reduce[0][0]']   
                                                                                                  
 block6g_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6g_activation[0][0]',  
 y)                                                                  'block6g_se_expand[0][0]']   
                                                                                                  
 block6g_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6g_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6g_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6g_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6g_drop (Dropout)      (None, 7, 7, 192)            0         ['block6g_project_bn[0][0]']  
                                                                                                  
 block6g_add (Add)           (None, 7, 7, 192)            0         ['block6g_drop[0][0]',        
                                                                     'block6f_add[0][0]']         
                                                                                                  
 block6h_expand_conv (Conv2  (None, 7, 7, 1152)           221184    ['block6g_add[0][0]']         
 D)                                                                                               
                                                                                                  
 block6h_expand_bn (BatchNo  (None, 7, 7, 1152)           4608      ['block6h_expand_conv[0][0]'] 
 rmalization)                                                                                     
                                                                                                  
 block6h_expand_activation   (None, 7, 7, 1152)           0         ['block6h_expand_bn[0][0]']   
 (Activation)                                                                                     
                                                                                                  
 block6h_dwconv2 (Depthwise  (None, 7, 7, 1152)           10368     ['block6h_expand_activation[0]
 Conv2D)                                                            [0]']                         
                                                                                                  
 block6h_bn (BatchNormaliza  (None, 7, 7, 1152)           4608      ['block6h_dwconv2[0][0]']     
 tion)                                                                                            
                                                                                                  
 block6h_activation (Activa  (None, 7, 7, 1152)           0         ['block6h_bn[0][0]']          
 tion)                                                                                            
                                                                                                  
 block6h_se_squeeze (Global  (None, 1152)                 0         ['block6h_activation[0][0]']  
 AveragePooling2D)                                                                                
                                                                                                  
 block6h_se_reshape (Reshap  (None, 1, 1, 1152)           0         ['block6h_se_squeeze[0][0]']  
 e)                                                                                               
                                                                                                  
 block6h_se_reduce (Conv2D)  (None, 1, 1, 48)             55344     ['block6h_se_reshape[0][0]']  
                                                                                                  
 block6h_se_expand (Conv2D)  (None, 1, 1, 1152)           56448     ['block6h_se_reduce[0][0]']   
                                                                                                  
 block6h_se_excite (Multipl  (None, 7, 7, 1152)           0         ['block6h_activation[0][0]',  
 y)                                                                  'block6h_se_expand[0][0]']   
                                                                                                  
 block6h_project_conv (Conv  (None, 7, 7, 192)            221184    ['block6h_se_excite[0][0]']   
 2D)                                                                                              
                                                                                                  
 block6h_project_bn (BatchN  (None, 7, 7, 192)            768       ['block6h_project_conv[0][0]']
 ormalization)                                                                                    
                                                                                                  
 block6h_drop (Dropout)      (None, 7, 7, 192)            0         ['block6h_project_bn[0][0]']  
                                                                                                  
 block6h_add (Add)           (None, 7, 7, 192)            0         ['block6h_drop[0][0]',        
                                                                     'block6g_add[0][0]']         
                                                                                                  
 top_conv (Conv2D)           (None, 7, 7, 1280)           245760    ['block6h_add[0][0]']         
                                                                                                  
 top_bn (BatchNormalization  (None, 7, 7, 1280)           5120      ['top_conv[0][0]']            
 )                                                                                                
                                                                                                  
 top_activation (Activation  (None, 7, 7, 1280)           0         ['top_bn[0][0]']              
 )                                                                                                
                                                                                                  
 global_average_pooling2d_1  (None, 1280)                 0         ['top_activation[0][0]']      
 1 (GlobalAveragePooling2D)                                                                       
                                                                                                  
 global_max_pooling2d_10 (G  (None, 1280)                 0         ['top_activation[0][0]']      
 lobalMaxPooling2D)                                                                               
                                                                                                  
 concatenate_9 (Concatenate  (None, 2560)                 0         ['global_average_pooling2d_11[
 )                                                                  0][0]',                       
                                                                     'global_max_pooling2d_10[0][0
                                                                    ]']                           
                                                                                                  
 dropout_45 (Dropout)        (None, 2560)                 0         ['concatenate_9[0][0]']       
                                                                                                  
 dense_19 (Dense)            (None, 12)                   30732     ['dropout_45[0][0]']          
                                                                                                  
==================================================================================================
Total params: 5950044 (22.70 MB)
Trainable params: 5889436 (22.47 MB)
Non-trainable params: 60608 (236.75 KB)
__________________________________________________________________________________________________

2023-11-26 20:33:15.205783: W tensorflow/core/util/tensor_slice_reader.cc:98] Could not open models/efficintB0_cw_imagenet: FAILED_PRECONDITION: models/efficintB0_cw_imagenet; Is a directory: perhaps your file is in a different file format and you need to use a different restore operator?

30/30 [==============================] - 6s 155ms/step


# Mostrar classification report
print("Classification Report EfficientNetV2 B0 (ImageNet Weights)")
print_classification_report(df_dataset_valid["label"].to_numpy(), prediction)

Classification Report EfficientNetV2 B0 (ImageNet Weights)
Classification Report:
               precision    recall  f1-score   support

           0       0.88      0.79      0.83       131
           1       1.00      0.98      0.99       123
           2       0.95      0.98      0.97       104
           3       1.00      0.99      0.99       100
           4       0.95      0.95      0.95        95
           5       0.96      0.97      0.97        78
           6       0.95      0.95      0.95        77
           7       1.00      0.93      0.96        58
           8       0.52      0.60      0.56        53
           9       0.96      0.98      0.97        47
          10       0.81      0.96      0.88        45
          11       0.98      1.00      0.99        45

    accuracy                           0.92       956
   macro avg       0.91      0.92      0.92       956
weighted avg       0.93      0.92      0.92       956

	path	label
0	dataset/train/Scentless Mayweed/8c496e84a.png	2
1	dataset/train/Common wheat/4a56f32c6.png	10
4	dataset/train/Fat Hen/9708f9c0e.png	4
5	dataset/train/Sugar beet/6d623072a.png	6
8	dataset/train/Small-flowered Cranesbill/869252...	3
...	...	...
5534	dataset/train/Scentless Mayweed/948251df3.png	2
5535	dataset/train/Sugar beet/29a0e6bf9.png	6
5537	dataset/train/Charlock/8b35222d0.png	5
5539	dataset/train/Shepherds Purse/953ced7c6.png	9
5543	dataset/train/Cleavers/b6220c08e.png	7

	path	label
3	dataset/train/Black-grass/e0380dff9.png	8
10	dataset/train/Scentless Mayweed/d748c7307.png	2
15	dataset/train/Fat Hen/e6b756e98.png	4
22	dataset/train/Small-flowered Cranesbill/ecf58a...	3
24	dataset/train/Charlock/d1b362c43.png	5
...	...	...
5500	dataset/train/Loose Silky-bent/fc2b27fff.png	0
5501	dataset/train/Charlock/fc3e15a2e.png	5
5503	dataset/train/Scentless Mayweed/d1e775b97.png	2
5505	dataset/train/Common Chickweed/f50c8181a.png	1
5538	dataset/train/Maize/e30accd2f.png	11

	path	label	split
0	dataset/train/Scentless Mayweed/8c496e84a.png	2	Train
1	dataset/train/Common wheat/4a56f32c6.png	10	Train
2	dataset/test/99569b224.png	Unkown	Test
3	dataset/train/Black-grass/e0380dff9.png	8	Valid
4	dataset/train/Fat Hen/9708f9c0e.png	4	Train
...	...	...	...
5539	dataset/train/Shepherds Purse/953ced7c6.png	9	Train
5540	dataset/test/8cf909eb3.png	Unkown	Test
5541	dataset/test/98062cd87.png	Unkown	Test
5542	dataset/test/cc3d2a59a.png	Unkown	Test
5543	dataset/train/Cleavers/b6220c08e.png	7	Train

	path	label
2	dataset/test/99569b224.png	Unkown
6	dataset/test/d14aa43f3.png	Unkown
7	dataset/test/b47691c08.png	Unkown
12	dataset/test/19fdf19fb.png	Unkown
14	dataset/test/7d4cd07ad.png	Unkown
...	...	...
5526	dataset/test/b30ab4659.png	Unkown
5536	dataset/test/bb1c84bbc.png	Unkown
5540	dataset/test/8cf909eb3.png	Unkown
5541	dataset/test/98062cd87.png	Unkown
5542	dataset/test/cc3d2a59a.png	Unkown

Redes Neuronales y Deep Learning: Proyecto de programación "Deep Vision in classification tasks"¶

ID Group: 07MIAR06¶

Dataset: Plant Seedlings Classification¶

Integrantes:¶

Álvaro Sánchez Rocamora¶

Daniel Abellán Sánchez¶

Carlos Melgosa Durá¶

El código de la práctica se encuentra en el siguiente repositorio de GitHub: https://github.com/daniabellan/Plant-Seedlings-Classification¶

Disclaimer:¶

Imports Librerías¶

Descarga de datos desde la plataforma Kaggle¶

Para este proyecto se va a hacer uso del dataset de kaggle https://www.kaggle.com/c/plant-seedlings-classification/data¶

En esta ocasión, por dificultades en la descarga directa del dataset desde kaggle, hemos obtado por descargarlo y hacer uso de éste en local.¶

Dentro del repositorio de Github con el código, en la carpeta de dataset/ disponemo de dos subcarpetas llamadas test/ y train/ con las muestras pertinentes a cada partición. Todo esto es respetando la jerarquía original del dataset.¶

1. Carga del conjunto de datos¶

Definimos las funciones necesarias para la carga de los datasets¶

Definimos las funciones para la creación del Datset de TensorFlow¶

Cargamos el dataset en un pd.DataFrame y visualizamos los datos¶

En el DataFrame anterior podemos visualizar por cada índice, el path de la imagen junto a su etiqueta asociada y al split que perteneces (Train, Valid o Test).¶

Aquellas imagenes pertenecientes a Test presentan la etiqueta de Unkown. Esto se debe a que el dataset de kaggle no tiene el dataset de test etiquetado, por ello para diferenciarlo aún más le ponemos esta etiqueta.¶

2. Inspección del conjunto de datos¶

Dado que ya disponemos del DataFrame con todas las muestras, vamos a crear cada partición y a su vez, visualizar las distribuciones de las etiquetas por cada uno de los splits.¶

Si nos fijamos en la distribución de muestras por clases del splir de train podemos ver como se presenta un desbalanceo de clases notable. Podemos ver como de la clase 0 disponemos de 523 muestras, mientras que de la clase 11 disponemos nomás de 176 muestras.¶

Dado este desbalanceo, se va a obtar por entrenar los modelos pasándoles los pesos de cada una de las clases mediante el netodo compute_class_weight() de la librería de sklearn.¶

Otra opción era duplicar las muestras de las clases minoritarias e igualar las muestras de la mayoritaria, pero esto supondría la repetición de muchas imágenes iguales y se ha descartado¶

Por lo que para calcular los pesos de las clases, que posteriormente se les pasarán en el fit del modelo, se hace uso del siguiente código.¶

Del resultado podemos ver como las clases minoritarias presentan mayor peso que las que tienen mas presencia en nuestro dataset. Por ello, de las clases con mayor peso, la red ponderará más.¶

Este cálculo de los pesos sólo lo haremos para el set de Train, para los otros no es necesario¶

Vistas las distribuciones de muestras por Split, vamos a visualizar unas muestras del Entrenamiento.¶

3. Acondicionamiento del conjunto de datos¶

Creados los Splits para el Dataset original, podemos pasar a preparar un ejemplo de los TensorFlow Datasets pertinentes para cada uno de los Splits (Train, Val y Test) que se van a usar para las siguientes pruebas. A su vez, también los visualizaremos para ver qué imagenes entrarían a la red¶

4.1 Desarrollo de la arquitectura de red neuronal y entrenamiento de la solución (from scratch)¶

Creamos los callbacks¶

Creamos los datasets para train, val y test¶

Función para lanzar factoría de modelos from scratch¶

5.1 Monitorización del proceso de entrenamiento para la toma de decisiones (from scratch)¶

Lanzar entrenamiento from scratch¶

5.2 Monitorización del proceso de entrenamiento para la toma de decisiones (pre-entrenada)¶

DenseNet¶

Obtenemos métricas para el dataset de validación¶

VGG16¶

Obtenemos métricas para el dataset de validación¶

EfficientNetV2 B0 (ImageNet Weights)¶

Obtenemos métricas para el dataset de validación¶

6.1 Evaluación del modelo predictivo y planteamiento de la siguiente prueba experimental (from scratch)¶

Estrategia 1: Entrenar desde cero o from scratch¶

Hiperparámetros de entrenamiento:¶

Top model:¶

Class Weights:¶

Model1:¶

Capas convolucionales:¶

Técnicas de regularización:¶

Model1_nobatchnorm:¶

Capas convolucionales:¶

Técnicas de regularización:¶

Model2:¶

Capas convolucionales:¶

Técnicas de regularización:¶

Model2_nobatchnorm:¶

Capas convolucionales:¶

Técnicas de regularización:¶

Model3:¶

Capas convolucionales:¶

Técnicas de regularización:¶

Model3_nobatchnorm:¶

Capas convolucionales:¶

Técnicas de regularización:¶

Análisis de resultados:¶

Model1:¶

Model1 No Batch Normalization:¶

Model2:¶

Model2 No Batch Normalization:¶

Model3:¶

Model3 No Batch Normalization:¶

6.2 Evaluación del modelo predictivo y planteamiento de la siguiente prueba experimental (pre-entrenada)¶

DenseNet121:¶

VGG16:¶

EfficientNetV2 B0:¶

**Redes Neuronales y Deep Learning: Proyecto de programación "Deep Vision in classification tasks"**¶

El código de la práctica se encuentra en el siguiente repositorio de GitHub: https://github.com/daniabellan/Plant-Seedlings-Classification ¶

Para este proyecto se va a hacer uso del dataset de kaggle https://www.kaggle.com/c/plant-seedlings-classification/data ¶